<![CDATA[Abstrak.Tingginya prevalensi stunting dipicu oleh kurangnya kualitas hidup balita di awal pertumbuhannya. Hal ini dapat berpengaruh pada rendahnya kualitas sumberdaya manusia dari banyak generasi penerus bangsa. Kajian stunting secara spasial menggunakan artificial neural network (ANN) bertujuan untuk mengetahui pola spasial dan prediksi tingkat kerawanan di wilayah lain di sekitarnya. Analisis dilakukan berdasarkan kondisi sosial-ekonomi dan budaya dari orang tua balita penderita stunting yang diperoleh dari wawancara, diolah dengan inverse distance weighted (IDW) dan diintegrasikan dengan hasil olah citra satelit Landsat 8 OLI-TIRS, berupa percent building density (PBD), land surface temperature (LST), normalized difference water index (NDWI), soil adjusted vegetation index (SAVI), dan normalized difference built-up index (NDBI). Model ANN dijalankan dengan metode back propagation, variasi jumlah hidden layer sebanyak 3, 5, dan 7, dengan variasi input prediksi mampu menghasilkan variasi distribusi stunting dan tingkat akurasinya. Berdasarkan nilai root mean square error (RMSE), bertambahnya jumlah hidden layer dan variasi input prediksi berkontribusi untuk menghasilkan akurasi hasil prediksi lebih baik, yakni 68%-93%. Secara spasial, keduanya secara langsung menjelaskan juga perubahan distribusi pola spasial kerawanan stunting di keseluruhan wilayah studi. Abstract. Lower toddler's life quality triggers the high prevalence of stunting at the beginning of their growth. This factor can affect many future generations' low quality of human resources. Studying stunting spatially using an artificial neural network (ANN) aims to determine the spatial pattern and predict the level of vulnerability in other surrounding areas. The analysis was carried out based on the socio-economic and cultural conditions of parents of children with stunting obtained from interviews, processed by inverse distance weighted (IDW) and integrated with the results of Landsat 8 OLI-TIRS satellite imagery, in the form of percent building density (PBD), land surface temperature (LST), normalized difference water index (NDWI), soil adjusted vegetation index (SAVI), and normalized difference built-up index (NDBI). The ANN model is run using the back propagation method, with variations in the number of hidden layers as many as 3, 5, and 7, with variations in predictive input capable of producing variations in the stunting distribution and the level of accuracy. Based on the value of the root mean square error (RMSE), the increasing number of hidden layers and variations in input predictions contribute to producing better prediction accuracy, which is 68%-93%. Spatially, both directly explain the changes in the distribution of the spatial pattern of stunting susceptibility in the entire study area. ]]>
