Application of data mining techniques for predicting rice crop yield in semi-arid climatic zone of India

The process of developing knowledge from the use of large data sets as an input and extracting useful information as an output is referred to as data mining. This acquired knowledge can be further applied by domain experts for decision making. In present research data mining techniques were applied to the historical agricultural dataset of semi-arid climatic zone of India to extract knowledge for predicting rice crop yield of kharif season. Free and open source software WEKA (Waikato Environment for Knowledge Analysis) was used to apply data mining techniques for the present agricultural dataset. Sensitivity, specificity and accuracy were computed to validate the experimental results. F1 score was computed to measure the test's accuracy. MCC (Mathews Correlation Coefficient) and was used to measure the quality of classification. Mean absolute error (MAE), root mean squared error (RMSE), relative absolute error (RAE) and root relative squared error (RRSE) were also calculated. The study found that J48 and LADTree classifiers provided the best performance among the classifiers used for the semi-arid climatic zone of India data set.