Optimasi Multi-Layer Perceptron Neural Network Menggunakan Particle Swarm Optimization dan Genetic Algorithm untuk Memprediksi Ketahanan Pangan di Wilayah Sumatera Utara dalam Mendukung Pembangunan yang Berkelanjutan

Abstract

According to the Global Food Security Index(GFSI) 2022,Indonesia scored 60.2 ranking 63rd out of 113 countries, below the average of several Asian nations in food security.The main challenges include a malnutrition rate of 6.5% and a stunting prevalence of 31.8%, while Law No. 18 of 2012 concerning food security has not been fully achieved. This study aims to develop a Multi Layer Perceptron(MLP) model optimized with Particle Swarm Optimization (PSO) and Genetic Algorithm(GA) to predict food security in North Sumatra. The food security prediction serves as an early warning system to detect potential food vulnerabilities.The MLP model is optimized due to its tendency to get trapped in local minima and its less efficient architectural design, which can impact model performance. Optimization is performed by searching for the best hyperparameters usingPSO and GA. The results indicate that optimizing the MLP hyperparameters with GA yields a Mean Squared Error(MSE) of 479.59 and a training time of 152.62seconds, whichis better than the PSO results of MSE 779.96 and a training time of 174.92 seconds. This optimization effectively improves prediction accuracy and training efficiency, makingthe model more effective in predictingfood security.