A Predicative Knowledge Based Irrigation Decision Through Artificial Neural Networks (ANN)

Abstract

Knowledge is the key for human thoughts and decision making. Developing a knowledge based management system(KBS) to mimic human thought is a critical task in a way of capturing the knowledge. Knowledge is generated from the experience and take long time to learn, also domain based. In Agriculture specially in farm irrigation, the selection of proper irrigation system plays vital role in sustainable crop production by taking attention of scarcity of water. The selection of irrigation method depends on four dimensional areas which are soil, water atmosphere and crop. Traditionally in on-farm water management, the decision marking is based on the farmer’s perspective and mostly leads to over irrigation or under irrigation. In this article, we trying to evaluate the performance of Artificial Neural Network (ANN) to capture the knowledge from various history of irrigation record and methods which are include fifteen irrigation parameters and four irrigation methods such as drip, sprinkler, border and furrow irrigation. The parameters are analysed and recorded in the data sheet with respective decision for ANN process. ANN is the key for used to classify the classes based on the attributes, therefore The five ANN classifier were selected for this study which are multilayer perceptron, support vector machine (SVM), J48 IBK and naive Bayesian. Thorough this study, the above five ANN classifiers were evaluated based on their performance by implementing the algorithm for predicative knowledge based to the irrigation decision. In the beginning of the study, more than twenty-five parameters were selected as attributes then they were reduced as sixteen considering weightage contribution to accuracy in irrigation decision. Based on the study, the multilayer perceptron shows the better performance with 99.7299 correctly identified instances as Irrigation decision, K = 0.996, MAE=0.0014 and RMSE=0.0186 than other four classifiers. Finally, the multilayer perceptron was selected to model the Predicative Knowledge Based System for Irrigation decision.