A web-based decision support system for evaluating soil suitability for cassava cultivation

Abstract

Precision agriculture in recent times had assumed a different dimension in order to improve on the poor standard of agriculture. Similarly, the upsurge in technological advancement, most especially in the aspect of machine learning and artificial intelligence, is a promising trend towards a positive solution to this problem. Therefore, this research work presents a decision support system for analyzing and mining knowledge from soil data with respect to its suitability for cassava cultivation. Past data consisting of some major soil attributes were obtained from relevant literature sources. This data was preprocessed using the ARFF Converter, available in WEKA. 70% of the data were used as training data set while remaining 30% were used for testing. Classification rule mining was carried out using J48 decision tree algorithm for the data training. ‘If-then’ construct models were then generated from the decision tree, which was used to develop a system for predicting the suitability status of soil for cassava cultivation. The percentage accuracy of the data classification was 76.5% and 23.5% for correctly classified and incorrectly classified instances respectively. Practically, the developed system was esteemed a prospective tool for farmers, soil laboratories and other users in predicting soil suitability for cassava cultivation.