A Genetic Programming Study on Classification of Cassava Plant

Abstract

Cassava (Manihot esculenta Crantz) is an important plant that is consumed in many forms. It could be processed as vegetable, chips, fodder, or bioethanol through a fermentation process. The cyclic acid HCN of cassava varies based on the varieties. Cassava with high HCN is toxic when it is consumed directly. This research designed a system to identify the cassava varieties based on HCN content by applying a heuristic search algorithm, using genetic operations. The identification of HCN content by applying Generic programming produced a structured classification rule and represents in tree form. The experiment in this study used binary code data generated from booleanizing process. Binary code data is divided into training data and test data using 5-fold cross-validation, and then the process of genetic operation. Rules are derived from repeated experiments to get the best rule. The best rule to identify with an average accuracy of 95.26%, obtained on population parameters of 10,000, 20-30 nodes. The node consists of Function set of AND, OR, NOR and 96 terminal sets (attributes / identifiers); in addition, the best classification rules are obtained on the crossover probability of 0.9 and 0.1 mutations of 10 generations. The resulting Rule can be utilized by the community in identifying the class of HCN cassava content.