Hits : 5
Genetic Programming for Medicinal Plant Family Identification System
Indra Laksmana1, Yeni Herdiyeni2 & Ervizal A.M. Zuhud3
1Mejor of Agricultural Machinery and Equipment, Payakumbuh Agricultural Polytechnic, Jalan Raya Negara Km 7, TanjungPati, West Sumatra, Indonesia 26271 2Department of Computer Science, Faculty of Mathematics and Natural Sciences Bogor Agricultural University, Darmaga Campus,Meranti Street, Wing 20 Level 5, West Java, Indonesia 16680 3Department of Forest Resources Conservation and Ecotourism, Faculty of Forestry, Bogor Agricultural University, Darmaga campus Lingkar Akademik, West Java, Indonesia 16680 Email: indra@politanipyk.ac.id
Abstract. Information about medicinal plants that is available in text documents is generally quite easy to access, however, one needs some efforts to use it. This research was aimed at utilizing crucial information taken from a text document to identify the family of several species of medicinal plants using a heuristic approach, i.e. genetic programming. Each of the species has its unique features. The genetic program puts the characteristics or special features of each family into a tree form. There are a number of processes involved in the investigated method, i.e. data acquisition, booleanization, grouping of training and test data, evaluation, and analysis. The genetic program uses a training process to select the best individual, initializes a generate-rule process to create several individuals and then executes a fitness evaluation. The next procedure is a genetic operation process, which consists of tournament selection to choose the best individual based on a fitness value, the crossover operation and the mutation operation. These operations have the purpose of complementing the individual. The best individual acquired is the expected solution, which is a rule for classifying medicinal plants. This process produced three rules, one for each plant family, displaying a feature structure that distinguishes each of the families from each other. The genetic program then used these rules to identify the medicinal plants, achieving an average accuracy of 86.47%.
Keywords: Booleanization; fitness evaluation; function set; genetic programming; medicinal plant families; terminal set.
Download Article
|