Text Mining for Pest and Disease Identification on Rice Farming with Interactive Text Messaging

Abstract

To overcome pests and diseases of rice farming, farmers always rely on information and knowledge from agricultural experts for decision making. The problem is that experts are not always available when the farmers need and the cost is quite high. Pests and diseases elimination is hard to be done individually since the farmers are lack of knowledge about the pest types that attack the rice fields. The objective of this study is to build a knowledge-based system that can identify pests and diseases interactively based on the information that has been told by the farmers using SMS communication services. The system can provide a convenience way to the farmers in delivering pests and disease problem information using a natural language. The text mining method performs tokenizing, filtering and porter stemming that used to extract important information sent by a SMS service. The method of Jaccard Similarity Coefficient (JSC) was used to calculate similarities of each pest and disease based on symptoms that are sent by the farmers through SMS. The corpus database usedin this study consists of 28.526 root words, 1.309 stop wordsand 180 words list. Pest and disease database reference in this study was obtained from the Ministry of Agriculture and Fisher (MAF) Timor-Leste. The result of the experiment shows that the system is able to identify the symptoms based on the keywords identified with the accuracy of 81%. The result of pest and disease identification has the accuracy of 86%.