What’s the Matter?

According to the FAO (Food and Agricultural Organization), food security can be achieved when everyone has sufficient and sustainable access to nutritious food. Good food security positively impacts various aspects of life, including community welfare. Among the different aspects of food security, food affordability is the main factor. Food crises still occur in various parts of the world, caused by wars and the COVID-19 pandemic, which restricted human movement. Globally, United Nations Development Programme (UNDP) reports show that the second Sustainable Development Goal (SDGs) (Zero Hunger) has not been maximally achieved in addressing hunger issues.

In Indonesia, food security is regulated by Law Number 18 of 2012 concerning Food, which contributes to national Sustainable Development Goals. According to the Global Food Security Index (GFSI) 2022 data from The Economist Group, Indonesia ranks 63rd out of 113 countries, indicating poor food security. The National Food Security Index (IKP) data indicates that Eastern Indonesia is a food-vulnerable area. Currently, the Indonesian government can only produce food security data at the district/city level once a year, while food security characteristics within districts/cities tend to be heterogeneous, making sub-district level data highly needed.

Although there is food security data at the sub-district level, the precision of this data is still low, as indicated by the high (Relative Standard Error) RSE value. A new framework has been proposed to improve data precision at the sub-district level by utilizing the Small Area Estimation (SAE) model. The implementation of geospatial big data also provides additional benefits in the form of spatial information and faster data updates compared to the current government data.

Alternative: A New Proposed Framework

A new framework has been proposed to enhance the precision of Indonesia's government food security data, particularly concerning food affordability. The sub-district level data on food affordability in East Nusa Tenggara Province for the year 2022 was used to implement this proposed framework. This data suffers from low precision and includes several non-sample areas, resulting in some sub-districts lacking estimation values. Therefore, in addition to working to improve data precision using the Small Area Estimation (SAE) model, this framework is capable of generating estimates by leveraging information from auxiliary variables.

In general, food affordability in Indonesia is measured using three dimensions: economic, social, and physical. The economic dimension is represented by the percentage of poverty, the social dimension by the percentage of households with a food share exceeding 65%, and the physical dimension by access to electricity. For each dimension, an SAE (Small Area Estimation) model will be developed using auxiliary variables derived from big geospatial data.

The innovation in this research lies in the implementation of composite formation proposed as a big geospatial data variable. The composite is created using three composite scenarios and four transformation scenarios. The three composite scenarios include the Weighted Sum Model method with Eigen Value as the weight, Geometric mean, and Arithmetic mean. Meanwhile, the four proposed transformation scenarios are min-max transformation, logarithmic transformation, Yeo-Johnson transformation, and untransformed data.

The best composite for each dimension will be selected using the All Possible Regression method, considering the R-square and AIC (Akaike Information Criterion) values from all twelve scenarios.

Two SAE models have been proposed to identify the best model: the Empirical Best Linear Unbiased Prediction (EBLUP) model and the Hierarchical Bayes Beta (HB-Beta) model. Auxiliary variables from the Village Potential data (PODES) and the population census (SP) are used as benchmarks against the composite big data dataset.

The results indicate that the HB-Beta model, using composite big geospatial data auxiliary variables, has a lower RSE (or Coefficien of Variation, CV) compared to direct estimation datasets and other models. This means that the precision of the HB-Beta model's estimation results is significantly better than that of the other models, indicating much smaller variation in estimator errors (better accuracy). The performance of the HB-Beta model is indicated by the blue line being below the other lines.

Based on the best model results for each dimension, the Spatial Food Affordability Index (SFAI) was formulated. The distribution of SFAI values in East Nusa Tenggara (NTT) Province is relatively good overall, but some areas are highly vulnerable, such as Mutis, Kodi Balaghar, Waiblama, Solor Selatan, Wewewa Utara, Palue, and Amofang Timur sub-districts. The average SFAI value in NTT is 74.55%, indicating a relatively high level of spatial food affordability.

For validation and evaluation, official data was used as a benchmark. The available official data at the sub-district level comprises indicators from the National Socioeconomic Survey (Susenas) related to food affordability. Pearson correlation tests were employed to examine the relationship and pattern between SFAI and the Food Affordability Aspect (AKP). As shown in figure below, there is a very strong correlation (0.9979) between SFAI and AKP at the sub-district level. The R-Squared value of 0.9958 indicates that 99.58% of the AKP can be explained by the SFAI.

Pros and Cons

The Spatial Food Affordability Index (SFAI) has several advantages:

1. High Precision: Using the Small Area Estimation (SAE) model, the SFAI demonstrates a relatively low Coefficient of Variation (CV) compared to direct estimates from official statistics, indicating lower error variation and higher precision without direct surveys.

2. Estimation for Non-sample Areas: SAE models can estimate values for non-sampled areas by leveraging information from surrounding regions and auxiliary variables.

3. Flexibility with Geospatial Big Data: Combining satellite imagery and other geospatial big data offers cost-effective, rapid, and extensive data accessibility, allowing for periodic and economical high-precision modelling.

4. Effective Use of Composites: Utilizing composites as auxiliary variables reduces the complexity of satellite and geospatial data, minimizing anomalies through dimensionality reduction.

However, integrating geospatial big data with SAE models requires expertise in big data processing and modelling, making it challenging for non-experts. Additionally, the estimated beta coefficients (????̂) from the model cannot be reused for future estimates or different regions, necessitating periodic re-modelling. This research contributes a framework for using geospatial big data in SAE models to achieve high-precision sub-district-level food affordability estimates more efficiently than direct surveys.

Conclusion

From the explanation above, it can be concluded that the proposed framework has successfully enhanced the precision of sub-district-level food affordability spatial based (SFAI) in East Nusa Tenggara Province. This improvement is indicated by the low Relative Standard Error (RSE) in the estimation results using the framework compared to direct estimation. The effectiveness of the SFAI is further supported by the estimation results being similar to direct estimates but with significantly better precision. Additionally, this framework can optimally update spatially-based food affordability data continuously, adapting to the availability of geospatial big data such as satellite imagery.

Tuberculosis: The Leading Cause of Respiratory Disease Death by Infectious Agents Worldwide

Tuberculosis (TBC) is the primary cause of respiratory disease mortality caused by infectious organisms worldwide, and it continues to provide a serious problem in public health. (WHO, 2021). TBC is caused by Mycobacterium tuberculosis, a harmful bacterium transferred through the air when an infected person coughs, sneezes, or speaks (WHO, 2021). However, in some cases, these bacteria can also spread to other organs such as the kidneys, bones, or brain through the bloodstream, causing a disease called extrapulmonary tuberculosis (CDC, 2020). This situation exacerbates the overall public health burden and hampers global efforts aimed at controlling and eradicating the disease.

The rapid and widespread spread of tuberculosis highlights the significance of a comprehensive and integrated strategy to its management. Prevention, early detection, effective treatment, and public education about the hazards and transmission of tuberculosis are crucial. Detailed, accurate, and thorough data-driven approaches are critical for monitoring illness development, evaluating intervention success, and making sound judgments about health-care program implementation. Consistent research and worldwide collaboration are also required to produce new, more effective TB treatments.

Tuberculosis is a disease that affects not just individuals but also the economy and society. People with tuberculosis often have lower work productivity, higher medical treatment costs, and social stigma, which can lead to social isolation. In countries with high TBC rates, such as Indonesia and India, the impact is more acute, affecting millions of families and communities. As a result, TB control initiatives must engage various sectors, including governments, health institutions, communities, and the commercial sector, to foster patient recovery and prevent the disease from spreading further.

Given that tuberculosis is still the primary cause of respiratory death caused by infectious organisms worldwide, increased attention and commitment from all parties is urgently required. Strong health policies, proper financial support, and innovation in medical research and technology must all be created and implemented to lessen the disease's burden. Such collaborative and sustained actions can combat the disease while also greatly improving the quality of life in global communities.

Tuberculosis Prevalence Conditions in Indonesia

The Data from the Ministry of Health (2021) in Indonesia show that, with 969,000 cases, Indonesia has the second-highest number of tuberculosis cases worldwide, behind India. Additionally, the Indonesian Ministry of Health projected that in 2022, there would be 52 cases of tuberculosis deaths for every 100,000 people (Ministry of Health, 2023). As a result, tuberculosis (TB) is a serious health problem that needs to be addressed seriously by everyone. Then, West Java and East Java provinces are the areas with the greatest number of cases.

Only district-or city-level aggregate data on tuberculosis cases in Indonesia is available, and it is updated annually by the Ministry of Health; hence, it lacks the specific details required for a more thorough examination. This restriction might result from things like traditional surveillance-based data reporting and collection techniques, which require a lot of money, time, and labor. Because of the disease's rapid spread, data processing and verification are crucial, as is specialized knowledge (Ministry of Health, RI, 2017).

However, as part of the National Medium-Term Development Plan (RPJMN) 2020–2024, the National Tuberculosis Control Strategy in Indonesia for the period 2020–2024 sets goals to lower the incidence of tuberculosis to 65 cases per 100,000 people by 2030 and the mortality rate from tuberculosis to 37 cases per 100,000 people by 2024 and 6 cases per 100,000 people by 2030 (Ministry of Health, RI, 2020).

In order to achieve this goal, a different method of presenting tuberculosis susceptibility risk data is required in order to enable the monitoring and assessment of complete health development planning. Additionally, more rapid updates and more detailed data presentation are required.

Our Innovation: Combination of Multisource Geospatial Big Data and Machine Learning Algorithm

The application of geospatial big data, encompassing remote sensing satellite imagery and points of interest (POIs), offers a transformative approach to overcome the limitations of official tuberculosis (TBC) data. Remote sensing involves acquiring information about the Earth's surface using electromagnetic radiation emitted from devices not in direct physical contact with the terrain (Statistics Indonesia-Badan Pusat Statistik, 2022). This technology encompasses a range of products derived from satellite sensors such as Suomi-NPP VIIRS, Sentinel 2 Multi Spectral Instrument (MSI) Level 2A, Sentinel-5 Precursor, Terra-MODIS, and NASA SRTM.

Moreover, points of interest (POIs) are geospatial datasets that include location-specific data about the surface of the Earth (Liu & Xiong, 2013). These datasets are derived from OpenStreetMap and other similar open-source platforms. The Meta Facebook Relative Wealth Index (Chi et al., 2022) is a noteworthy geospatial dataset that provides extensive coverage of wealth distribution data. Then, the Gridded Population from WorldPop provide detailed demographic data. The utilization of these data collection techniques yields several benefits, such as improved population characteristic insights, efficient and objective monitoring of socio-economic and physical phenomena, cost effective, and a more comprehensive representation of geographic areas (Putri et al., 2022; Ramadhan & Wijayanto, 2023; Utami et al., 2023).These geospatial datasets can be used to create precise maps that show the risk of tuberculosis susceptibility with the integration of machine learning techniques.

The Data and Variables Used

The data used in this research came from a variety of geospatial big data sources, including satellite images, Points of Interest (POIs), and other geospatial datasets. Satellite imagery data were acquired between January 1, 2022, and December 31, 2022, with a focus on East Java Province. The POI data was obtained from OpenStreetMap, while the Relative Wealth Index from Meta Facebook provided the other geospatial data.

The POI data, which totaled over 17,000 points, was divided into three categories: health, education, and economic facilities. The number of these amenities within each 1-kilometer resolution area was then counted to demonstrate infrastructural density. Furthermore, the distance from the center of each 1-kilometer resolution region to the nearest facility was determined using Euclidean distance to demonstrate accessibility. The Relative Wealth Index (RWI) is another geospatial big data source used to assess the local population's well-being. Statistics Indonesia (Badan Pusat Statistik, BPS) provided official total population statistics, while BPS collected TB prevalence data through the Indonesian Ministry of Finance. To estimate tuberculosis prevalence, the data were classified into several categories.

Air quality, green open space, population density, infrastructural density and accessibility, regional well-being, physical environment, and regional topography are among the aspects considered. Here is a visualization mapping a general overview of the aspect approach used using the multisource geospatial big data that has been described.

Our Methodology Proposed

The feature extraction results of those multisource geospatial big data features that have been feature engineered will become independent variables in each machine learning and deep learning algorithm. The final result is Tuberculosis Susceptibility Risk Area Estimation mapping at the best granularity level of 1 km x 1 km spatial resolution in East Java, Indonesia. The Web Map Development Dashboard is then created to make it easier for users, particularly associated stakeholders, to access research data and make better decisions and policies about health development.

A Granular Mapping of the Tuberculosis Susceptibility Risk Area with 1-Kilometer Resolution

A spatial granular estimate is essential for both identifying places with larger concentrations of deprivation and providing a through grasp of the risk distribution in order to map tuberculosis susceptibility risk areas efficiently. The use of sophisticated spatial analytic methods enables a more accurate and thorough evaluation of TB susceptibility.

Using multisource geospastial big data, this case study estimates tuberculosis (TBC) susceptibility levels in East Java. The physical environment, geographical topography, population density, infrastructural accessibility and density, air quality, and green open spaces are a few of these aspects. Artificial Intelligence (AI) techniques like machine learning and deep learning can be used to combine and analyze this data to estimate TB susceptibility at a granular level throughout the province.

The geographical resolution of a 1-kilometer area inside East Java Province enables a more accurate and focused estimation of tuberculosis susceptibility risk. With this degree of specificity, it is possible to identify communities and regions that are more likely to become infected, which makes it easier to build interventions and policies that are specifically designed to address the problems these populations experience in developing their health. The estimate process produces spatially explicit data, which is displayed through maps or other graphical representations. These maps offer important insights on how the risk of tuberculosis susceptibility varies throughout East Java Province's various regions. Stakeholders and legislators can use this information to pinpoint problem areas and put practical solutions in place to alleviate disparities in health and development.

Decision-makers can learn more about the localized dynamics of TBC susceptibility risk in East Java Province by carrying out a case study with a 1-kilometer area resolution. The development of customized initiatives, plans, and policies targeted at lowering inequality of health development and enhancing the wellbeing of the local populace can be guided by the information provided.

Our machine learning algorithm-developed mapping results demonstrate a very strong correlation: 74.9% of the official tuberculosis prevalence statistics at the regency/municipality level can be explained by a granular mapping of tuberculosis susceptibility risk. This highlights the advantages of using multisource geospatial big data and machine learning methods together, as we suggest, to enhance policymaking by supplying high-quality data as alternative data to support the official data.

One benefit of the 2-dimensional web map dashboards that we deploy is that they allow users—especially stakeholders—to identify precise locations inside each 1-kilometer area. The map dashboard created in this study includes a number of extra elements that improve stakeholders' and users' capacity to recognize tuberculosis susceptibility risk estimates and related indicators in a more precise, user-friendly, and interactive manner.

VISIT 2D MAP DASHBOARD HERE!

Health Development Priority Area for Preventing Tuberculosis Disease

Health development policies prioritize areas based on the granular mapping estimation of tuberculosis susceptibility risk at the finest 1 km x 1 km resolution. Using the distribution of the General Allocation Fund (DAU), this study illustrates the advantages of utilizing comprehensive geographic big data to coordinate development policy direction at the lowest administrative levels, such as sub-village or village. A web-based map dashboard designed specifically for this purpose facilitates policy review. Stakeholders are assisted in identifying priority development areas for policy execution by the map dashboard, which aggregates data from the village to the sub-district level of tuberculosis susceptibility risk mapping.

Please Provide Us With Your Advice and Feedback for Our Evaluation, Here!

Urgensi Ketersediaan Data Konektivitas dan Aksesibilitas Jalan

Jalan merupakan infrastruktur penting dalam kehidupan, terutama di bidang transportasi dan logistik. Jalan memberikan kemudahan akses dari suatu tempat ke tempat lainnya​.

United Nations, dalam Sustainable Development Goals (SDGs) telah menetapkan salah satu tujuan pembangunan berkelanjutan tepatnya pada target 9.1 yaitu mengembangkan infrastruktur yang berkualitas untuk akses yang terjangkau dan merata bagi semua, salah satu implementasinya adalah dengan pembangunan jalan.

Sayangnya, fakta di lapangan menunjukkan bahwa kepadatan jalan di Indonesia berada pada peringkat 93 dari 196 negara di dunia yaitu dan masih berada di bawah angka rata-rata kepadatan jalan dunia. Selain itu, Kementerian Pekerjaan Umum dan Perumahan Rakyat (PUPR) menyatakan bahwa rata-rata existing waktu tempuh jalan nasional di Indonesia masih berada di bawah negara tetangga seperti Thailand dan Malaysia (Kementerian PUPR, 2022).

Peran Penyediaan Data yang Granular dan Up-to-Date

1. Mendukung Indonesia dalam pembangunan sektor jalan dunia

Dalam rangka memperkuat kerjasama Association of Southeast Asian Nations (ASEAN), pada tahun 2015 negara anggota ASEAN sepakat untuk mewujudkan ASEAN Community dengan 3 (tiga) pilar yaitu: (i) Komunitas Politik - Keamanan ASEAN; (ii) Komunitas Ekonomi ASEAN; dan (iii) Komunitas Sosial - Budaya ASEAN.

Dalam konteks ini, konektivitas merupakan salah satu aspek terpenting dalam rangka mewujudkan visi dari Komunitas ASEAN tersebut, yang diterjemahkan dalam bentuk proyek Trans Asia - ASEAN Highways. Data aktual yang granular dibutuhkan untuk menganalaisis dan menentukan wilayah yang prospektif untuk pembangunan ASEAN Highways (Rencana Strategis Direktorat Jenderal Bina Marga 2020-2024).

2. Meningkatkan peranan konstruksi terhadap PDB

Badan Pusat Statistik (BPS) menyatakan bahwa jalan merupakan konstruksi yang paling berpengaruh terhadap Produk Domestik Bruto (BPS, 2023; Informasi Statistik Infrastruktur PUPR, 2022). Ketersediaan data yang semakin baik dapat membantu meningkatkan peranan konstruksi terhadap PDB.

3. Perdagangan Antar Wilayah

Pada tahun 2021, sebesar 77,56% perdagangan antar wilayah di Indonesia menggunakan angkutan darat sebagai moda transportasi utama (BPS, 2022). Selain itu Badan Pusat Statistik (BPS), Badan Perencanaan Pembangunan Nasional (Bappenas) dan UNFP memproyeksikan bahwa pada tahun 2025, tingkat urbanisasi nasional akan mencapai angka 60%. Ketersediaan data yang lebih granular dan komprehensif dapat menjadi dasar yang kuat bagi pemerintah dalam pembentukan kebijakan lingkungan strategis nasional yang tepat sasaran (Renstra DJBM 2020-2024).

4. Pengelolaan APBN oleh Kementerian PUPR

Berdasarkan Informasi Statistik Infrastruktur PUPR, Anggaran Pendapatan dan Belanja Negara (APBN) terbesar dialokasikan untuk program infrastruktur konektivitas yaitu mencapai 39,60%. Ketersediaan data geospasial yang aktual dengan pendekatan remote sensing akan mengurangi pendanaan untuk pengumpulan data.

Kondisi Data Konektivitas dan Aksesibilitas Jalan Saat ini

Saat ini, data infrastruktur jalan nasional di Indonesia menjadi tanggung jawab Kementerian PUPR, khususnya Direktorat Jenderal Bina Marga (DJBM) serta lembaga turunannya yaitu Balai Pelaksanaan Jalan Nasional (BPJN). Data konektivitas dan aksesibilitas jalan yang tersedia saat ini, diterbitkan di dalam Laporan Akuntabilitas Kinerja Instansi Pemerintah (LAKIP) DJBM yang diperbarui setiap satu tahun sekali. Pengumpulan data yang dilakukan masih menggunakan pendekatan konvensional yaitu dengan melakukan pengecekan langsung (ground check) ke lapangan. Pendekatan ini membutuhkan biaya dan waktu yang besar dalam pengimplementasiannya. Disamping itu, data konektivitas dan aksesibilitas jalan yang tersedia hanya sampai pada level nasional. Belum terdapat data yang aktual dan komprehensif terkait konektivitas dan aksesibilitas jalan yang dapat dijadikan pedoman untuk pembuatan kebijakan di level yang lebih granular.

Dengan mempertimbangkan aspek tersebut, diperlukan pembaharuan data geospasial jalan secara berkala dengan kuantitas sumber daya manusia, waktu dan biaya yang lebih rendah untuk memastikan bahwa data yang ada dapat mencerminkan kondisi aktual dari objek spasial yang bersangkutan serta dapat melakukan perhitungan indeks konektivitas dan aksesibilitas suatu wilayah yang lebih granular dan komprehensif.

Pemanfaatan Big Data Geospasial Multisumber

Saat ini, diketahui bahwa pemanfaatan big data geospasial multiusmber melalui teknologi remote sensing mampu menjadi solusi yang menjanjikan. Remote sensing adalah pendekatan yang mampu menggambarkan ciri khas suatu wilayah bumi dengan merekam spektrum elektromagnetik dari jarak jauh menggunakan satelit sehingga menghasilkan data citra satelit tanpa melakukan observasi langsung ke lapangan (J. B. Campbell and R. H. Wynne, 2011). Dibandingkan dengan pendekatan konvensional, remote sensing memberikan keunggulan yang lebih besar, seperti kemudahan pengambilan data, pembaruan data dalam jangka waktu yang lebih singkat, cakupan wilayah yang lebih granular, serta biaya yang lebih ekonomis bila dibandingkan dengan pendekatan konvensional.

Data dan Variabel yang Digunakan

Terdapat beberapa jenis big data geospasial yang akan digunakan dalam penelitian ini. Pertama, data citra satelit Sentinel-2. Kedua, data Night Time Light (NTL) dari NOAA-VIIRS. Ketiga, data Population Density dari WorldPop.

Metodologi

Hasil Pemetaan Indeks Geospasial Konektivitas dan Aksesibilitas Jalan

Dalam penelitian ini, digunakan pendekatan Multi-Criteria Decision Analysis (MCDA); Weighted Sum Model (WSM), Weighted Product Model (WPM), dan Višekriterijumsko Kompromisno Rangiranje/Multi-Criteria Compromise Ranking (VIKOR) dengan bobot berdasarkan korelasi pearson. Selanjutnya dilakukan evaluasi model dan dipilih model terbaik untuk melakukan pembangunan Indeks Geospasial Konektvitas dan Aksesibilitas Jalan yang dipetakan dalam bentuk peta hasil estimasi dan dashboard inetraktif pada level kecamatan.

Hasil pemetaan yang diperoleh menunjukkan bahwa 65,5% variasi dari IGKJ telah mampu dijelaskan oleh model. Sementara itu, 82,0% IGAJ telah mampu dijelaskan oleh model serta memberikan representasi yang akurat dari kondisi nyata di lapangan sehingga berpotensi dalam mendukung pembuatan kebijakan terkait infrastruktur jalan di Indonesia. Pemetaan hasil estimasi ini juga dapat dilihat lebih jelas pada dashboard 2D interaktif yang telah dibangun berikut.

KLIK UNTUK MENGUNJUNGI DASHBOARD IGKJ DAN IGAJ PROVINSI JAMBI 2023

Wilayah Prioritas untuk Perbaikan IGKJ dan IGAJ di Provinsi Jambi

Dari visualisasi estimasi IGKJ dan IGAJ diatas, terlihat bahwa wilayah permukiman dengan sebaran populasi yang padat cenderung memiliki konektivitas dan aksesibilitas jalan yang tinggi. Sementara itu, wilayah non-permukiman, cenderung memiliki konektivitas dan aksesibilitas jalan yang rendah. Berdasarkan hasil estimasi IGKJ dan IGAJ, dibentuk daftar kecamatan prioritas untuk perbaikan IGKJ dan IGAJ pada tabel berikut.

Permasalahan Sampah Rumah Tangga (SRT)

Jumlah penduduk yang kian hari kian bertambah tak lepas dari jumlah sampah rumah tangga (SRT) yang makin meningkat pula. Permasalahan ini juga menjadi isu internasional yang ditargetkan untuk segera diselesaikan. Salah satunya tertuang dalam SDGs 12.5 terkait mengurangi produksi limbah dengan indikatornya berupa jumlah timbulan sampah yang didaur ulang (Bappenas, 2020).

Di Indonesia, berdasarkan UU No. 18 Tahun 2008, sampah rumah tangga (SRT) adalah sampah yang berasal dari kegiatan sehari-hari dalam rumah tangga yang tidak termasuk tinja dan sampah spesifik. Berdasarkan situs Sistem Informasi Pengelolaan Sampah Nasional (SIPSN) yang dikelola oleh Kementerian Lingkungan Hidup dan Kehutanan (KLHK), selama tiga tahun terakhir, 2020-2022, rata-rata timbulan sampah di Indonesia terus mengalami peningkatan, yaitu di tahun 2022 timbulan sampah harian nasional mencapai 98.942,33 ton atau meningkat 30% dibandingkan tahun 2020. Menelisik lebih dalam, akar permasalahan ini ialah SRT yang menumpuk. Hal ini dapat terlihat pada Gambar 2 bahwa di tahun 2022 dominasi SRT dalam timbulan sampah ialah sebesar 40%

Ketersediaan Data

Dengan sampah yang terus menumpuk, upaya perbaikan tentunya diperlukan, terutama terkait pengelolaan SRT. Perbaikan yang tepat tak lepas dari rancangan strategi yang matang dengan berlandaskan data yang ada. Meski begitu, data SRT yang tersedia di Indonesia masih terbatas dengan tingkat penyajian paling kecil ialah kabupaten/kota. Padahal sejatinya, statistik yang lebih terpilah (disagregasi) dan terperinci (granular) dapat membantu memfasilitasi penargetan kebijakan atau intervensi yang lebih efisien (Asian Development Bank, 2020). Sejatinya, di Indonesia telah ditetapkan metode konvensional untuk menyediakan data SRT hingga wilayah yang lebih kecil, yaitu dengan pengambilan sampel berdasarkan SNI 19-3964-1994 (BSN, 1994). Meski begitu, metode konvensional yang ditawarkan tersebut memerlukan biaya yang tidak sedikit serta waktu yang lebih lama untuk pengambilan sampel sehingga metode konvensional juga memiliki keterbatasan berupa cakupannya yang tidak luas dan memerlukan sumber daya yang tidak sedikit. Oleh karena itu, pendekatan baru yang lebih efektif diperlukan untuk dapat mengestimasi SRT hingga wilayah yang lebih kecil sehingga strategi yang dibuat dapat lebih presisi atau tepat sasaran.

Pemanfaatan Potensi Data Alternatif: Integrasi Data Official Statistics dan Big Data Geospasial

Dengan perkembangan teknologi saat ini, pemanfaatan big data geospasial (citra satelit hasil pengindraan jauh, Point of Interest (POI), dan data lainnya) yang memiliki keunggulan mudah, cepat, murah, dan tersedia hingga wilayah granular berpeluang untuk dimanfaatkan dalam penyelesaian masalah ini. Tentunya dalam pemilihan potensi big data geospasial tersebut, variabel-variabel yang bersumber dari official statistics diperlukan untuk menjadi acuan sehingga pemilihan big data lebih terarah dan tepat sasaran. Oleh karena itu, integrasi pemanfaatan potensi data alternatif berupa big data geospasial dengan data official statistics diusulkan untuk menjadi solusi dari permasalahan ini. Berikut adalah data official statistics yang digunakan beserta indikator usulan big data geospasialnya.

1. Rasio Perkotaan

Rasio perkotaan didapatkan dengan perhitungan jumlah desa/kelurahan yang terklasifikasi perkotaan dibagi dengan seluruh jumlah desa/kelurahan. Data tersebut bersumber dari Publikasi Peraturan Kepala BPS No. 120 Tahun 2020 Tentang Klasifikasi Desa, Perkotaan dan Perdesaan di Indonesia 2020. Berikut adalah indikator usulan big data geospasial untuk variabel rasio perkotaan yang bersumber dari satelit Sentinel-2, Suomi NPP-VIIRS, WorldPop, dan Wilkesrtat BPS.

2. Rasio Membuang di Tempat Sampah

Rasio membuang di tempat sampah ialah rasio desa dengan sebagian besar keluarga membuang sampah ke tempat sampah kemudian diangkut yang bersumber dari pendataan potensi desa (Podes) 2021. Hal ini didasari oleh SRT yang tertimbang dan tercatat dalam SIPSN merupakan SRT yang diangkut atau terkelola dari fasilitas/sarana pengelolaan sampah, sesuai dengan Peraturan Menteri Lingkungan Hidup dan Kehutanan P.10/MENLHK/SETJEN/PLB.0/4/2018. Big data geospasial yang digunakan ialah Point Of Interest (POI) tempat pembuangan sampah yang bersumber dari integrasi data KLHK (Kementerian Lingkungan Hidup dan Kehutanan) serta Google Maps.

3. Sex Ratio

Variabel sex ratio ialah rasio perempuan per 100 laki-laki. Hal ini didasari oleh penelitian Wut et al (2021) yang menemukan bahwa responden perempuan lebih banyak melakukan perilaku pro lingkungan yang merujuk pada kesadaran pengelolaan sampah dibandingkan laki-laki dan hasil penelitian Talalaj & Walery (2015) menunjukkan korelasi yang kuat antara tingkat timbulan sampah per kapita tahunan dengan rasio perempuan terhadap laki-laki (r = 0,81). Berikut indikator usulan big data geospasial yang bersumber dari WorldPop.

Metodologi

Hasil Pemetaan Estimasi SRT Granular Level Grid 1 Km di Jawa Tengah

Dari hasil evaluasi model machine learning, model terbaik yang terbentuk ialah model CatBoost Regression (CBR). Model terbaik inilah yang digunakan untuk mengestimasi SRT level granular yang hasilnya divisualisasikan ke dalam map dashboard. Dalam visualisasi, terdapat 3 pengelompokan data SRT, yaitu rendah, sedang, dan tinggi.

• Rendah ialah nilai SRT yang dihasilkan berada pada rentang rata-rata Jawa Tengah ke bawah (<= 0,15 (kg/orang/hari))
• Sedang ialah nilai SRT berada pada rentang di atas rata-rata Jawa Tengah tetapi masih berada pada rata-rata nasional kebawah (0,15 - 0,21 (kg/orang/hari)).
• Tinggi ialah nilai SRT berada pada rentang di atas rata-rata nasional (> 0,21(kg/orang/hari))

Hasil pemetaan level grid 1 km ini dapat diagregasi pada wilayah administrasi yang ada, mulai dari desa, kecamatan, kabupaten/kota, hingga provinsi. Oleh karena itu, pemetaan ini dapat dimanfaatkan dalam menentukan wilayah prioritas dalam upaya perbaikan pengelolaan SRT di Jawa Tengah mulai dari tingkat grid dan wilayah administrasi terkecil.

Why is Human Development Index Data so Important and Strategic?

The Sustainable Development Goals (SDGs) are a global action plan developed by the United Nations (UN) to end poverty, protect the planet, and ensure that all people enjoy justice and prosperity (United Nations, 2015).

In today's modern world, quality of life and development lead to the improvement of human living standards and well-being as well (Sen, 1999). The importance of good human development means that everyone should have an equal opportunity to achieve this well-being (Haq, 1995). This is because the goal of human development is to create a society that has a long life, lives a healthy life, and leads a productive life in an equitable and sustainable manner (Statistics Indonesia -Badan Pusat Statistik, 2022). Thus, economic growth and progress are not the only way to measure the success of human development; the improvement of the quality and standard of living of the community is also an important indicator.

The quality of life and human development are measured by the Human Development Index (HDI), introduced by the United Nations Development Programme (UNDP) in 1990. The HDI is based on three dimensions: a long and healthy life, measured by Life Expectancy at Birth (LE); education, represented by Expected Years of Schooling (EYS) and Mean Years of Schooling (MYS); and a decent standard of living, assessed through gross national income per capita (In Indonesia is using adjusted real per capita expenditure) (Statistics Indonesia-Badan Pusat Statistik, 2023).

Those indicators align with the SDGs on health and well-being (Goal 3), quality education (Goal 4), and inclusive economic growth (Goal 8) (United Nations, 2016). HDI is crucial for evaluating human development and ranking regions or countries. In Indonesia, it is a strategic measure of government performance and is used to determine the General Allocation Fund (DAU) (Statistics Indonesia - Badan Pusat Statistik, 2022).

The Condition of HDI Achievement in Indonesia

In 2022, the HDI in Indonesia was 72.91 (Statistics Indonesia-Badan Pusat Statistik, 2023). On the other hand, the government targets the HDI in Indonesia to reach 75.54 in order to achieve equitable development and quality of human life in the Medium-Term Development Plan (RPJMN) target (The Ministry of National Development Planning, 2019). However, the level of official HDI data presentation in Indonesia is limited to the regency or municipality level, with updates once a year. In addition, the dominant source of HDI data collection comes from the conventional method, the National Socio-Economic Survey (SUSENAS), which requires much greater effort, cost, and time. Achieving this target requires granular data presentation and faster updates to assist in the monitoring and evaluation process of comprehensive human development planning, so an alternative method of presenting HDI data is needed.

Our Innovation: Combination of Geospatial Big Data and Machine Learning Algorithm

The use of geospatial big data, such as remote sensing satellite imagery and points of interest (POIs), have the potential to overcome the limitations of official HDI data. Remote sensing is a method for collecting information about the surface of the Earth through electromagnetic radiation produced by a device that is not in direct contact with the object (Statistics Indonesia-Badan Pusat Statistik, 2022). One of the products of remote sensing is sensor capture from satellite imagery such as Suomi-NPP VIIRS, Sentinel 2 Multi Spectral Instrument (MSI) Level 2A, Sentinel-5 Precursor, Terra-MODIS, and NASA SRTM spacecraft.

Additionally, POIs are geospatial datasets that contain information about location points on the Earth's surface (Liu & Xiong, 2013). Some other geospatial big data products are WorldPop's Gridded Population and Meta Facebook's Relative Wealth Index (Chi et al., 2022), which provide detailed data coverage. Some of the advantages of collecting data with this technique are: provide more accurate data or correspond to the true state of the population; objectively and effectively observe socio-economic and physical phenomena; lower costs; and provide a more granular representation of their territory (Putri et al., 2022; Ramadhan & Wijayanto, 2023; Utami et al., 2023). Then, by combining use of those geospatial big data with machine learning algorithms, it is possible to extract granular mappings of the quality of life and well-being indicators to estimate the HDI.

The Data and Variables Proposed

The satellite imagery from many sensors, POIs, and other geospatial big data sources are some of the sources of the data used. The satellite image data was collected between January 1, 2022, and December 31, 2022, with a locus in South Sumatra Province. The road network map data was sourced from OpenStreetMap, while the POIs data were obtained from Google Maps.

The crawled POIs data is a total of 92,371 points grouped by the categories of health, education, and economic facilities. A total of 7,707 are educational facilities consisting of Primary School (4,831), Junior High School (1,254), Senior High School (1,188), and Univerity/College (434). Meanwhile, there are 4,809 health facilities consisting of hospitals (154), Community Health Center (1,124), Clinic/Medical Centre/Doctor (1.797), and Pharmacies/Drug Store (1,014). The largest number is economic facilities, which is 80,575 consisting of Public Services (28,355), Trade & Business (18,460), Finance and Commercial (12,258), Recreation & Tourism (9,265), Transportation (6,702), and Communication (5,535). We then counted the number of such facilities within each 1-kilometer resolution area to illustrate infrastructure density, and calculated the distance of each central 1-kilometer resolution area to the nearest facility using haversine distance to illustrate accessibility. To measure an infrastructure's connection to the transport system we also calculated the distance of each facility to its nearest road network from OpenStreetMap using the same Haversine distance.

The Relative Wealth Index (RWI), High Resolution Population Density (derived from Meta Facebook), and Gridded Population (supplied from WorldPop) are other geospatial big data sources that are utilized. The official statistics data from the 2021 Village Potential Data (PODES) recording and the overall population were then obtained from Statistics Indonesia (BPS) as a supplementary dataset. In order to estimate indicators on the HDI, each data used is categorized into aspects.

Our Methodology Framework

The feature extraction results of geospatial big data variables that have been feature engineered will become independent variables in each machine learning model. In addition, the Spatial Granularity Human Development Index (SGHDI) is also developed for HDI mapping through the approach of each indicator. The final result of both is HDI mapping at the best granularity level of 1 km x 1 km spatial resolution.

The Spatial Granularity Mapping of Human Development Index with High Resolution

In order to map the human development index, spatial granular estimation is essential for both identifying areas with higher concentrations of deprivation and offering a comprehensive understanding of the index's distribution. A finer-grained assessment of human development can be obtained by applying sophisticated spatial analytic tools.

This case study estimates South Sumatra of the human development levels using a variety of data sources. Environmental factors, socioeconomic indicators, demographic data, geographical datasets, and other pertinent local data are some of these data sources to each dimensions. Spatial modeling approaches can be used to estimate human development throughout the province by combining and analyzing this data.

Within South Sumatra Province, a more precise and targeted calculation of the human development index is possible because to the spatial resolution of a 1-kilometer area. It makes it possible to identify groups and regions that are more susceptible to poor development, which makes it easier to establish tailored interventions and policies to address the particular difficulties that people in such areas experience.

Spatially explicit data, which can be seen through maps or other graphical representations, are the results of the estimate process. These maps offer insightful information about how the human development index is distributed among the various regions in the South Sumatra Province, which can help stakeholders and policymakers identify areas of concern and put into place practical solutions to reduce development inequality.

Decision-makers can learn more about the localized dynamics of human development in South Sumatra Province by carrying out a case study with a 1-kilometer area resolution. This information can direct the creation of customized plans, initiatives, and regulations aimed at lowering inequality and enhancing the welfare of the populace in the region.

The mapping results of our proposed Spatial Granularity Human Development Index explain 73.8% of the official HDI data at the regency/municipality level and are highly correlated at the indicator level. Mapping with the help of machine learning algorithms is very likely to exceed our SGHDI, which can explain 71.8% of official data at the sub-district level and 85.2% at the regency/municipality and both are very strongly correlated. This proves the advantages of using the combination of geopastial big data and machine learning algorithms that we propose to assist policy making through the provision of quality data.

The 2-dimensional and 3-dimensional map dashboards have the advantage of allowing users to pinpoint specific geographical places within each 1km area. The map dashboard produced in this research includes various additional components that can help stakeholders and users to identify HDI estimates and each of its indicators more accurately, more intuitively, and interactively.

VISIT 2D MAP DASHBOARD HERE!

VISIT 3D MAP DASHBOARD HERE!

Human Development Priority Areas

The results of HDI granular mapping estimation at the best 1 km x 1 km level can be used as the basis of priority areas for human development policy direction. The results of the HDI granular estimation at the 1 km level in this study show the advantages of detailed areas that can be covered from the use of geospatial big data, but the direction of development policies is carried out at the smallest level, namely sub-village/village using the distribution of the General Allocation Fund (DAU). The policy evaluation can be identified through the web map dashboard that has been built. HDI estimation figures at the granular level have been aggregated from the village level to the sub-district level on the map dashboard so that it can facilitate stakeholders in determining the policy of development priority areas.

There are many areas with HDI achievements below the national average (national HDI = 72.91), namely 3,119 urban villages/villages, and below the average of South Sumatra Province (South Sumatra HDI = 70.90), namely 2,382 urban villages/villages, based on the results of identifying the estimated HDI figures in 3,264 urban villages/villages through the map dashboard. To increase population welfare and meet the HDI target in the RPJMN IV agenda, these villages must be assessed and closely watched for greater human development in terms of the health, education, and economics of their citizens.

Pentingnya Ketahanan Pangan

Menurut UU 18 tahun 2012 tentang pangan, ketahanan pangan adalah kondisi terpenuhinya pangan bagi negara sampai dengan perseorangan, yang tercermin dari tersedianya pangan yang cukup, baik jumlah maupun mutunya aman, beragam, bergizi, merata, dan terjangkau, serta tidak bertentangan dengan agama., keyakinan, dan budaya masyarakat, untuk dapat hidup sehat, aktif, dan produktif secara berkelanjutan.
Dalam Sustainable Development Goals (SDGs) poin kedua, yaitu yaitu menghilangkan kelaparan (zero hunger), diharapkan segala bentuk kelaparan di dunia dapat berakhir, sehingga tiap individu berhak mendapatkan akses pangan yang layak (Vitriana, dkk., 2022). Ketahanan pangan yang kuat sangat diperlukan sebagai upaya untuk membentuk sumber daya manusia yang sehat, aktif dan produktif, sesuai dengan tujuan akhir dari pembangunan ketahanan pangan nasional oleh Food Agriculture Organization (FAO), yaitu membangun manusia-manusia yang bisa hidup dengan kualitas hidup yang lebih baik. Ketahanan pangan juga merupakan salah satu faktor penting dalam pembangunan ekonomi berkelanjutan (Shi & Xiao, 2022). Jika keadaan pangan suatu negara buruk maka akan menimbulkan berbagai masalah kesehatan yang juga akan berdampak pada munculnya masalah ekonomi.

Indeks Ketahanan Pangan

Untuk mengetahui tingkat ketahanan pangan suatu wilayah serta faktor-faktor pendukungnya telah dikembangkan beberapa indikator dalam bentuk indeks ketahanan pangan (Zamrodah, 2020). Di Indonesia, data resmi indeks ketahanan pangan dikelola dan dipublikasikan oleh Badan Ketahanan Pangan, Kementrian Pertanian yang dirilis tiap tahun dengan level terkecil penyajian data adalah kabupaten/kota

Data yang digunakan untuk menyusun indeks ketahanan pangan berasal dari berbagai kementerian/lembaga, yaitu kementrian pertanian, Badan Pusat Statistik (BPS), dan kementrian kesehatan. Pengumpulan data yang digunakan menggunakan survei sampel yang memiliki berbagai keterbatasan, seperti biaya yang mahal, cakupan yang terbatas, serta membutuhkan waktu yang lama untuk dapat menyajikan data tersebut. Akibatnya, data resmi indeks ketahanan pangan di Indonesia hanya dapat dirilis tiap tahun dengan level terkecil adalah kabupaten/kota. Sementara, pemerintah harus bisa mengeluarkan kebijakan yang sesuai dalam menangani bahkan mencegah kerentanan pangan di suatu wilayah. Oleh karena itu, dibutuhkan alternatif data yang dapat menggambarkan indeks ketahanan pangan yang memiliki cakupan yang lebih granular, dapat diperbaharui lebih cepat dengan biaya yang lebih rendah sehingga dapat digunakan untuk memantau indeks ketahanan pangan yang lebih granular dan up to date untuk mendukung pengambilan kebijakan mengenai ketahanan pangan yang lebih akurat.

Usulan Data Alternatif

Ketersediaan informasi dari data citra satelit dan big data geospasial dapat memberikan solusi atas keterbatasan data indeks ketahanan pangan yang ada. Berikut data yang digunakan dalam membangun estimasi indeks ketahanan pangan granular.

Metodologi

Hasil Pemetaan Estimasi Indeks Ketahanan Pangan Granular Level Grid 2.5 Km

Dalam penelitian ini didapatkan model Random Forest Regression (RFR) sebagai model terbaik untuk melakukan estimasi indeks ketahanan pangan yang selanjutnya diimplementasikan ke dalam map dashboard.

Urgensi Akses Air Minum Layak

Setiap manusia membutuhkan air. Menjaga kebersihan, mencuci, menyiapkan makanan, minum, dan sanitasi membutuhkan antara 20 dan 100 liter air setiap hari (Howard et al., 2020). Kondisi air yang dikonsumsi harus layak atau tidak tercemar. Kurangnya akses terhadap air minum dan sanitasi layak dapat menyebabkan berbagai penyakit seperti pneumonia, anemia, diare kronis, infeksi saluran cerna, dan lainnya (United  Nations Children’s Fund (UNICEF), 2008). Hampir dua miliar orang di seluruh dunia tidak memiliki akses ke air minum yang memadai (World Health Organization (WHO), 2019).

Penuntasan masalah akses terhadap air minum layak diangkat dalam Sustainable Development Goals (SDGs) poin 6.1:

"Pada tahun 2030, seluruh penduduk dunia memiliki akses yang universal dan merata terhadap air minum yang aman dan terjangkau".

Peran Penyediaan Data yang Granular dan Up-to-Date

Untuk mempercepat peningkatan akses air minum layak di Indonesia, data yang granular dan up-to-date sangat berperan dalam proses perencanaan dan pemantauan akses air minum layak.

1. Proyek Strategis Nasional

Peraturan Presiden Republik Indonesia Nomor 18 Tahun 2020 Tentang Rencana Pembangunan Jangka Menengah Nasional Tahun 2019-2024 menetapkan indikasi pendanaan sebesar Rp 123,4 triliun untuk pembangunan infrastruktur air leding perpipaan untuk 10 juta sambungan rumah tangga. Data aktual dengan georeferensi atau lebih terperinci hingga tingkat rumah tangga diperlukan untuk menentukan rumah tangga yang paling membutuhkan akses air minum (Ajisegiri et al., 2019).

2. Dana Desa

Menurut Peraturan Menteri Desa, Pembangunan Daerah Tertinggal, dan Transmigrasi No. 8 Tahun 2022, pemerintah desa membutuhkan data yang lebih rinci untuk mengevaluasi pengelolaan air minum masyarakat desa. Ketersediaan data di tingkat desa dapat membantu dalam perencanaan dan evaluasi penggunaan dana desa.

3. Rencana Aksi Daerah

Pemerintah kabupaten/kota juga harus melakukan evaluasi dan pemantauan realisasi anggaran berdasarkan data setidaknya dua kali setahun untuk melaksanakan Rencana Aksi Daerah Air Minum dan Pengelolaan Lingkungan (RAD AMPL).

Kondisi Data Saat Ini

Saat ini, data official tentang akses air minum layak diperoleh dari Survei Sosial Ekonomi Nasional (SUSENAS). Metode ini tidak hanya membutuhkan banyak sumber daya dan biaya, tetapi juga hanya dapat tersedia di tingkat kabupaten/kota dan diupdate hanya sekali setahun. Oleh karena itu, sulit untuk mengevaluasi Rencana Aksi Daerah Air Minum dan Pengelolaan Lingkungan dengan data SUSENAS dan menentukan lokasi mana yang paling penting untuk akses air minum.

Pemanfaatan Big Data Geospasial Multisumber

Sebaliknya, big data geospasial multisumber memiliki keunggulan dalam pengumpulan data yang cepat, mudah, murah, dan memberikan tingkat penyajian yang lebih granular, serta frekuensi update yang lebih cepat. Dengan demikian, big data geospasial multisumber dapat berfungsi sebagai sumber data alternatif untuk melengkapi keterbatasan data SUSENAS.

Data geospasial yang memiliki elemen kewilayahan seperti koordinat, alamat, kode pos, dan lainnya yang sangat besar atau melebihi kapasitas sistem komputasi individual saat ini disebut sebagai big data geospasial (Lee & Kang, 2015). Big data geospasial termasuk citra satelit hasil penginderaan jauh, Unmanned Aerial Vehicle/drone, point of interest, mobile positioning device, dan lainnya.

Data yang Digunakan

Pada penelitian ini menggunakan variabel big data geospasial dari berbagai jenis data seperti citra satelit, point of interest (POI), dan big data geospasial lainnya yaitu: relative wealth index.

Metodologi

Hasil Pemetaan Akses Air Minum Layak

Dalam kajian ini didapatkan dua pendekatan dalam memetakan akses air minum layak yaitu pendekatan machine learning dan multi-criteria decision analysis yang diimplementasikan ke dalam map dashboard.

Berikut merupakan map dashboard yang dihasilkan beserta form evaluasi dashboard.

Desa Prioritas

Dengan menggunakan model machine learning yang sama, dilakukan updating menggunakan data pada tahun 2022. Selanjutnya, didapatkan grid-grid pada desa yang memerlukan perhatian lebih dalam meningkatkan akses air minum layak. Berikut ditampilkan daftar desa prioritas peningkatan akses air minum layak di Provinsi Jawa Barat.

Evaluasi Dashboard

evaluasi

Introduction

The Importance of Sea Transportation

Sea transportation is one of the modes that are quite helpful for human activities, especially in distributing logistics (K. Verawati et al., 2022)

All types of activities from sea transportation should be monitored so that illegal activities do not occur that can harm one party. However, in detecting this marine vessel, an instrument that has good accuracy is needed to carry out fast information processing of marine transportation detection, especially this ocean liner (J. Wu et al., 2021).

Model Detection

Object Detection Model to Detect Marine Vessels Activities

In connection with the importance of detecting ship objects as one of the main modes of sea transportation, especially in Indonesia, technology or models are needed that can easily detect certain objects, in this case, they are objects of sea vessels. Redmon, et al. (J. Redmon et al., 2016) have made preliminary instructions on a model that can be used to detect objects, which is named YOLO (You Only Look Once), where the YOLO model is a model with one-step detection, where the model will be more efficient to carry out detection when compared to two-step detection such as RCNN, Faster RCNN, and the like which require two stages in detecting the main object to be detected (S. Luo et al., 2022) so that in this study the YOLO model will be used as a basis for detecting marine objects that are the main objects in their detection.

Model Architecture

YOLOv5x6 Architecture

A special architecture of the YOLOv5 version of 6.0, where this model has a higher level of complexity compared to other YOLOv5 derivatives. In the picture, it is clear that in the backbone, there is an addition of a CSP (Cross-Stage-Partial) function in the section, followed by SPPF (Spatial Pyramid Pooling-Faster) on the neck of the model, and also in this section, the concept of the Darknet-53 is used as a derivative of the YOLOv4 model (A. Bochkovskiy et al., 2020). With this SPPF, the size of the image received by the model can vary, and later when the training process is carried out, size adjustments will be made according to the parameters that have been set at the time of model construction. After that, with Conv2d, the model will detect objects at the image input according to a very wide range of points, namely 10 x 10, 20 x 20, 40 x 40, and 80 x 80.

Dashboard

Based on Streamlit Framework

This dashboard is deploy under the Streamlit server

Publication

Research results have been published on: https://ieeexplore.ieee.org/document/9994340

This paper get the best paper predicate in the conference

Evaluate Our Work!

For research purposes, users are expected to provide an evaluation of their experience using this prediction dashboard.

The 2022 UN Big Data Hackathon

Sahabat Data, Statmin mau berbagai info menarik nih. Sahabat Data pasti pernah mendengar istilah marathon atau pun triathlon, tapi apakah sudah pernah tahu dengan istilah Hakacthon??? Apa itu termasuk ke dalam cabang olahraga???

The United Nations Committee of Experts on Big Data and Data Science for Official Statistics (UNCEBD), Major Group for Children and Youth (MGCY) dan Badan Pusat Statistik (BPS) akan menyelenggarakan event besar bertajuk "2022 UN Big Data Hackathon". Acaranya akan berlangsung di Yogyakarta pada 8-11 November 2022.

Untuk mendapatkan info lebih lanjut, sahabat data bisa akses halaman https://unstats.un.org/bigdata/events/2022/hackathon/.

Why is Child Poverty a Crucial Concern?

The world has made remarkable progress in advancing development in recent years. However, the stark reality remains that more than 700 million people still live in extreme poverty (UNICEF, 2020). Among the most vulnerable and affected by this dire situation are children. Growing up in poverty, they often endure inadequate access to food, sanitation, housing, healthcare, and education, hindering their ability to survive and thrive.

UNICEF reports that approximately 1 billion children worldwide are deprived across multiple dimensions, lacking fundamental necessities such as proper nutrition and clean water. Shockingly, around 365 million children live in conditions of extreme poverty (UNICEF, 2020). In other words, over half of the global population enduring extreme poverty comprises children. This aligns with the World Bank's assertion that children face a higher risk of poverty compared to adults, primarily due to poor households tending to have more children than non-poor households (World Bank, 2011).

In the context of Indonesia, a developing nation, child poverty presents a critical challenge. According to data from the Central Statistics Agency in 2017, approximately 40% of Indonesia's impoverished population consists of children (Central Statistics Agency, 2017). Alarming trends reveal that impoverished children in Indonesia often remain trapped in poverty as they reach adulthood (Rizky, Suryadarma, & Suryahadi, 2019). These cycles of poverty perpetuate as the children of impoverished individuals are likely to be born into impoverished households. If left unaddressed, this persistent issue will perpetuate a never-ending cycle of poverty in Indonesia. Hence, it becomes imperative to prioritize the alleviation of child poverty as a means to tackle the larger challenge of sustained poverty in the country.

Child Poverty Alleviation

Child poverty is neither inevitable nor impervious to intervention. As evidenced by numerous countries, it can be mitigated and even eradicated through sustained attention and proactive measures (Edmonds, E. V., & Schady, N., 2012).

The Sustainable Development Goals (SDGs) have united nations in a historic commitment to address extreme child poverty. The SDGs strive to create a world where all children have the necessary resources to survive, thrive, and reach their full potential by reducing multidimensional child poverty by half before 2030, surpassing mere income-based measurements (UNICEF, 2020).

Indonesia, too, has embraced the SDG agenda on child poverty within its National Medium-Term Development Plan (RPJMN) spanning from 2015-2019 to the current period of 2020-2024 (Statistics, 2017). The 2020-2024 RPJMN of Indonesia has set forth comprehensive development targets to tackle child poverty at a macro level.

The Indonesian government has undertaken various initiatives to combat child poverty, including the implementation of programs such as the Program Indonesia Pintar (PIP), Program Keluarga Harapan (PKH), School Operational Assistance (BOS), and other targeted social assistance schemes. However, effective monitoring by policymakers is crucial to ensure the accurate and efficient delivery of poverty alleviation aid to the intended beneficiaries (Fiscal Policy Agency, 2021).

Utilization of Satellite Imagery and Other Geospatial Big Data

Geospatial Big Data refers to a vast collection of data that is geographically referenced. It encompasses a wide range of information that can be leveraged to understand various aspects, including economic activities and accessibility within a specific geographical area.

Satellite imagery is an integral part of geospatial big data, capturing measurements of objects and phenomena on the Earth's surface without physical contact. It provides a comprehensive view of the social and economic conditions in a given region, offering objective insights based on its distinct characteristics (Yao, Li, et al., 2017).

One of the major advantages of geospatial big data is its frequent and timely updates, providing a more up-to-date picture of the analyzed area. Additionally, its granular level of spatial detail enables a deeper understanding of localized patterns and trends. Furthermore, the utilization of geospatial big data is often more cost-effective compared to traditional data collection methods such as surveys or censuses, making it a compelling alternative for supporting data on child poverty (Pokhriyal, Zambrano, Linares, & Hernández, 2020).

By harnessing the power of geospatial big data, stakeholders such as policymakers, researchers, and organizations can gain valuable insights into the spatial dimensions of child poverty. This knowledge can inform targeted interventions, resource allocation, and the evaluation of poverty alleviation programs, ultimately leading to more effective measures in addressing child poverty.

Data Utilized

The data used is sourced from geospatial big data, specifically multi-source satellite imagery and other geospatial big data such as Point of Interest (POI) data.

Multi-source satellite imagery involves the integration of data from various satellite sources, providing a comprehensive view of the Earth's surface. These satellite images capture valuable information about land cover, land use, vegetation, and other spatial features. By combining data from multiple sources, it enhances the accuracy and richness of the information extracted.

In addition to satellite imagery, Point of Interest (POI) data is another valuable source within geospatial big data. POI data includes information about specific locations such as businesses, schools, healthcare facilities, and other relevant points of interest. It provides insights into the availability and accessibility of essential services for children living in different areas.

By utilizing multi-source satellite imagery and other geospatial big data such as POI data, analysts and researchers can gain a comprehensive understanding of the spatial dimensions of child poverty. These data sources enable the identification of areas with limited access to crucial resources and services, helping policymakers develop targeted interventions and allocate resources effectively.

The integration of geospatial big data enhances the analysis and visualization of child poverty, allowing for a deeper understanding of the spatial distribution and disparities. It provides valuable insights into the physical environment, infrastructure, and socio-economic factors that influence child well-being.

Overall, leveraging multi-source satellite imagery and geospatial big data, including POI data, supports evidence-based decision-making and enables more effective strategies in addressing child poverty by identifying areas in need and directing interventions to where they are most needed.

Methodology

Spatial Granular Estimation for Mapping Child Poverty

Spatial granular estimation plays a crucial role in mapping child poverty, providing a detailed understanding of its distribution and identifying areas with higher concentrations of vulnerable children. By utilizing advanced spatial analysis techniques, it becomes possible to estimate child poverty at a fine-grained level.

In this case study, various data sources are utilized to estimate child poverty levels in Banten Province. These data sources include socioeconomic indicators, demographic information, geospatial datasets, and other relevant data at a local level. By integrating and analyzing these data, spatial modeling techniques can be applied to estimate child poverty across the province.

The spatial resolution of a 1-kilometer grid allows for a more detailed and localized estimation of child poverty within Banten Province. It enables the identification of specific areas and communities that are more vulnerable to poverty, facilitating targeted interventions and policy development to address the unique challenges faced by children in those areas.

The output of the estimation process will be spatially explicit data, which can be visualized through maps or other graphical representations. These maps provide valuable insights into the distribution of child poverty across different regions within Banten Province, aiding policymakers and stakeholders in identifying priority areas and implementing effective measures to alleviate child poverty.

By conducting a case study with a 1-kilometer grid resolution in Banten Province, decision-makers can gain a deeper understanding of the localized dynamics of child poverty. This knowledge can guide the development of tailored strategies, programs, and policies to reduce child poverty and improve the well-being of children in the province.

Evaluate Our Work!

For research purposes, users are expected to fill out a user survey regarding their experience using this dashboard.

With satellite imagery and point of interest (POI), poverty can be identified from space!

Poverty as One of SDGs' Main Goals

Considering that poverty is still a problem in many countries, “poverty alleviation” in 2030 will become the “backbone” of the goals of the sustainable development agenda (SDGs).

Poverty which is the main goal of the MDGs is again the main goal in the SDGs. Apart from the fact that poverty and hunger are still global problems, the elimination of poverty as the main goal is directed at ensuring the sustainability of the MDGs achievements.

Poverty is defined as an economic inability to meet basic needs (food, housing, clothing, education, and health) as measured by the expenditure approach (Statistics Indonesia, 2021).

Statistics Indonesia, locally known as Badan Pusat Statistik (BPS), recorded that the number of poor people in Indonesia in September 2021 amounted to around 26.5 million people, with a percentage of the national poor population is 9.71%.

Remote Sensing Satellite Imagery and Point of Interest (POI)

Remote sensing satellite imagery and Point of Interest (POI) have the ability to show human activity from a geographical aspect so that it can be used as an approach to describe poverty (Shi et al, 2020).

Remote sensing is a method of capturing the characteristics of an area of the earth using remote electromagnetic observations collected through satellite (Campbell & Wynne, 2011).

POI is a geodataset containing information on the location of important points on the earth's surface that can be obtained through OpenStreetMap (OSM) with a very fast update period. This data can provide information related to urban structures and economic facilities (Yao, Li, et al., 2017).

Compared to conventional socio-economic survey data, remote sensing and POI satellite image data is a valuable resource (Triscowati et al., 2020, 2019; Wijayanto et al., 2020) because of its uniqueness and objectivity in viewing the socio-economic and physical phenomena of the environment with multi-scale perspective effectively and accurately (Shi et al., 2020).

Data Used in This Study

The data used are multisource satellite imagery from remote sensing and POI from OpenStreetMap!

1. NTL (Nighttime Light Intensity) to detect the distribution of economic activity at night
2. Normalized Difference Vegetation Index (NDVI) to detect rural areas by vegetation
3. Built-Up Index (BUI) to detect urban areas by building distribution
4. Normalized Difference Water Index (NDWI) for land cover detection
5. Land Surface Temperature (LST) for urban areas detection,  according to the distribution of the land surface temperature
6. Carbon Monoxide (CO) for economic activity detection based on vehicle and urban pollution
7. Nitrogen Dioxide (NO2) for economic activity detection based on industry
8. Sulfur Dioxide (SO2) for economic activity detection based on energy consumption
9. POI Density to show the density of economic activity of a region
10. POI Distance to show the accessibility of a region

Grid Level 1.5 km Poverty Mapping

From multi-source satellite imagery and POI, monetary poverty estimation map was built using a deep learning model, and relative spatial poverty (RSPI) map was built using a weighted sum model as illustrated below.

Map Dashboard

The results obtained are then presented on an interactive dashboard map.

Publication

Research results have been published on: https://www.mdpi.com/2220-9964/11/5/275

Evaluate Our Work!

For research purposes, users are expected to provide an evaluation of their experience using this dashboard map.

Bertempat di Ruang Rapat Gedung 1 Lantai 2 Politeknik Statistika STIS, telah dilaksanakan knowledge sharing dengan topik "Opinion Mining-Citizen Feedback Data to Local Government Decision Making".

Knowledge sharing kali ini menghadirkan Dr. Wayne Wobcke dari School of Computer Science and Engineering, University of New South Wales sebagai narasumber.

Discussion Material

Dalam rangka meningkatkan wawasan dan menambah jejaring, Tim Big Data STIS melakukan studi banding ke beberapa tempat yang dilaksanakan pada tanggal 21-23 Oktober 2019 di Bandung. Ada tiga tempat yang dikunjungi yaitu:

1. Telkom University
2. Institut Teknologi Bandung
3. Lembaga Ilmu Pengetahuan Indonesia (LIPI)

Laporan Kunjungan

Telkom University

Dalam kunjungan pertama, tim berkunjung ke Social Computing and Big Data Laboratory yang merupakan bagian dari Digital Business Research Center yang ada di Telkom University. Tim disambut dengan hangat oleh Bapak Andry Alamsyah selaku ketua Research Center dan beberapa anggota timnya. Kunjungan ini dibuka dengan perkenalan melalui paparan profil instansi secara umum dan secara khusus profil dari tim peneliti di masing-masing instansi beserta penelitian-penelitian yang sudah dan sedang dilakukan oleh masing-masing tim.

Dari perkenalan tersebut Bapak Andry Alamsyah memaparkan sejarah dan proses terbentuknya Digital Business Research Center. Dikatakan bahwa Social Computing and Big Data Laboratory sudah didirikan sejak tahun 2012 namun baru pada tahun 2018 Research Center tersebut secara resmi muncul. Pada awal berdirinya, kegiatan yang dilakukan masih difokuskan untuk menghasilkan publikasi ilmiah sebanyak-banyaknya. Salah satu hasil penelitian, yang sebagian besar dilakukan dengan menggunakan data dari media sosial, yang telah dipublikasikan adalah Pendeteksian Objek Fashion Populer menggunakan Convolutional Neural Network.

Setelah itu, kegiatan lebih difokuskan untuk menghasilkan produk-produk, dimana produk yang sedang populer untuk saat ini adalah dalam bentuk training dan workshop untuk Artificial Intelligence. Dari kegiatan training tersebut peserta dapat belajar dan nantinya mendapat sertifikat pelatihan. Pak Andry mengatakan bahwa tenaga pengajar yang memberikan training tidak harus berasal dari internal sehingga terbuka peluang untuk melakukan kerja sama. Tidak hanya itu, salah satu kegiatan yang baru-baru ini mereka lakukan adalah Big Data for Teens dimana di dalamnya dilakukan edukasi kepada remaja mengenai data yang dihasilkan dari media sosial.

Statistik ofisial dan Big Data dapat berjalan bersama-sama, dimana statistik ofisial menghasilkan angka sementara Big Data digunakan untuk melihat trend atau pola yang dapat memberikan penjelasan lebih lanjut terhadap angka tersebut

Dalam diskusi lebih lanjut terlihat bahwa salah satu perbedaan antara Research Center mereka dengan Tim Big Data STIS adalah pada tujuan akhir yang ingin dicapai. Sebagai perguruan tinggi kedinasan yang berada di bawah naungan Badan Pusat Statistik, Tim Big Data STIS dapat menghasilkan laporan-laporan pendukung yang berasal dari Big Data untuk kemudian dapat disandingkan dengan statistik ofisial yang dikeluarkan BPS sehingga ada insight yang lebih di dalamnya. Pak Andry mengatakan bahwa memang statistik ofisial dan Big Data dapat berjalan bersama-sama, dimana statistik ofisial menghasilkan angka sementara Big Data digunakan untuk melihat trend atau pola yang dapat memberikan penjelasan lebih lanjut terhadap angka tersebut.

Institut Teknologi Bandung

Di hari berikutnya, Tim Big Data STIS mengunjungi Laboratorium Parallel Computing di Institut Teknologi Bandung. Di sana, tim bertemu dengan Bapak Achmad Imam Kistijantoro selaku Ketua Program Studi Sarjana Sistem dan Teknologi Informasi yang juga merupakan seorang ahli dalam hal Distributed Systems.

Dalam paparannya beliau mengatakan bahwa Big Data memiliki setidaknya dua jenis spesialisasi di dalamnya yaitu Data Engineer dan Data Analyst. Keduanya memiliki fokus dan disiplin ilmu yang berbeda yang nantinya bisa dikolaborasikan untuk membentuk suatu sistem yang berjalan dengan baik. Data Engineer lebih berfokus kepada bagaimana menyiapkan infrastruktur yang baik untuk mengambil, menyimpan dan mengolah data. Berbekal dari pengalaman, beliau juga menyampaikan terkadang untuk mengatasi suatu masalah dalam infrastruktur data, pendekatan non engineering juga perlu dilakukan. Sebagai contoh, dengan membuat SOP penjadwalan pengiriman data, arus masuk data dapat lebih terkontrol dan menjamin infrastruktur yang ada tetap available.

Dalam pengimplementasian infrastruktur data, beliau juga menekankan terhadap penggunaan program atau teknologi berbasis open source untuk setiap komponen. Penggunaan open source akan lebih menguntungkan dibandingkan dengan mengembangkan program atau teknologi sendiri dari awal mengingat open source dikembangkan secara kolaboratif oleh banyak pihak.

Lembaga Ilmu Pengetahuan Indonesia (LIPI)

Terakhir, Tim Big Data STIS bertemu dengan kelompok penelitian Machine Learning Pusat Penelitian Informatika, Lembaga Ilmu Pengetahuan Indonesia (LIPI). Seperti sebelumnya, kunjungan dibuka dengan perkenalan melalui paparan profil instansi secara umum beserta penelitian-penelitian yang sudah dan sedang dilakukan oleh masing-masing tim. Selain kelompok penelitian Machine Learning, di dalam Pusat Penelitian Informatika LIPI juga terdapat tiga kelompok penelitian lainnya yaitu kelompok penelitian Data Science, High Processing Computer, dan Computer Vision. Selanjutnya Bapak Hilman Ferdinandus Pardede memberikan pemaparan singkat terkait asal mula berkembangnya Artificial Intelligence.

Beberapa penelitian yang dilakukan di dalam kelompok penelitian ini antara lain adalah pengembangan POS tagger untuk Bahasa Indonesia, pendeteksian penyakit pada tanaman menggunakan foto, dan speech enhacement untuk secure voice communication. Dalam diskusi disebutkan bahwa salah satu tantangan yang dihadapi oleh tim peneliti di LIPI adalah pada tahapan pengumpulan data karena kegiatan tersebut dapat memakan waktu yang lama dan mengeluarkan biaya yang cukup banyak. Hal ini tentu membuka kepada peluang untuk dilakukannya joint research mengingat Tim Big Data STIS sebagai tim di perguruan tinggi kedinasan yang berada di bawah naungan Badan Pusat Statistik memiliki akses data yang lebih mudah. Penelitian yang lebih mendalam selanjutnya dapat dilakukan oleh tim peneliti dari LIPI. Tak hanya itu, kerja sama juga dapat dilakukan untuk penggunaan infrastruktur pengolahan data mengingat sudah tersedianya komputer yang memiliki tenaga komputasi yang tinggi di LIPI.

Kesimpulan

Dari kegiatan studi banding yang telah dilakukan, ada banyak poin menarik yang dapat diambil sebagai pelajaran untuk lebih berkembangnya kemampuan baik teknis maupun non teknis dari Tim Big Data STIS. Beberapa poin tersebut antara lain:

1. Penggunaan Big Data dan statistik ofisial dapat dilakukan secara beriringan dimana Big Data digunakan untuk melihat trend sementara statistik ofisial digunakan untuk memberikan angka statistik
2. Big Data memiliki setidaknya dua jenis spesialisasi di dalamnya yaitu Data Engineer dan Data Analyst. Keduanya memiliki fokus dan disiplin ilmu yang berbeda yang nantinya bisa dikolaborasikan untuk membentuk suatu sistem yang berjalan dengan baik
3. Pemanfaatan tools yang sudah tersedia khususnya yang berlisensi open source akan lebih efektif dibandingkan dengan mengembangkan tools sendiri. Kontribusi dapat dilakukan dengan mengembangkan tools tersebut lalu mengirimkan hasilnya ke komunitas terkait untuk di review lebih lanjut
4. Beberapa peluang kerja sama dapat dilakukan antar tim antara lain untuk pemasaran produk baik progam, penelitian maupun workshop dengan Telkom University dan juga melalui joint research dengan LIPI