Application of Artificial Neural Network in the Early Warning System for Food Crisis Management

Abstract

Artificial neural network (ANN) has widely used to various sectors in agriculture. In term of food security management, ANN used to determine food crisis level based on its factors. The aim of this research is to increase ANN performance in term of pattern recognition by advanced learning using updated data as well as ANN weight analysis. This research has used multi-layer perceptron 2 hidden layers with backpropagation algorithm. The input-output patterns were food crisis factors and crisis level, respectively. Result showed that advance learning could increase accuracy level. It was from 70,55% to 85,38%. Based on weight analysis of ANN neuron, factors that affected to crisis level were: (1) crop failure/natural disaster, (2) normative consumption ratio, (3) rice price, (4) stock exchange, (5) infant mortality, (6) non forest area, (7) currency, (8) people under poverty line, (9) underweight infant and (10) annual rainfall. The 3 big factors are critical aspect should be concerned in food crisis management. Keywords: ANN, backpropagation, food crisis management, food security