Enhancing the productivity of irrigated rice fields in West Nusa Tenggara through utilizing Multilayer Perceptron (MLP) and Self-Organising Maps (SOM)

Abstract

As Indonesia's population grows, ensuring a stable food supply becomes increasingly important. Recent changes in weather patterns have significantly impacted food production, particularly rice farming. In West Nusa Tenggara (NTB), a key area for rice production, maintaining consistent output is crucial. However, varying responses to unpredictable weather have led to significant differences in productivity across NTB's regencies and cities. This study aims to enhance the productivity of irrigated rice fields in NTB by predicting productivity levels for 2023 to 2024 using the best multilayer perceptron (MLP) model. We will compare 5 MLP model architectures to identify the optimal model for the prediction process. We will use the prediction results to cluster areas regionally through the self-organizing map (SOM) algorithm. We used the Davies-Bouldin Index (DBI) to determine the optimal number of clusters. This research compared DBI values for cluster counts of 2, 3, 4, and 5, determining the optimal cluster number by the smallest DBI value. The lowest DBI is 0.391 observed for 3 clusters. From this clustering, Cluster 1 consists of 7 regencies/cities with the lowest productivity level, Cluster 2 contains 1 regency with a moderate productivity level, and Cluster 3 includes 2 regencies/cities with the highest productivity level. The study concludes that the 7 regencies/cities in Cluster 1, identified as having low productivity require greater focus from local governments to optimize land area and paddy yields to enhance productivity in those areas.