COMPARATIVE STUDY OF MACHINE LEARNING ALGORITHMS FOR RAINFALL PREDICTION – A CASE STUDY IN NEPAL

One bitter truth associated with Nepal is it one of the richest country in fresh water resources in the world and also an agriculture dominant country. However, it largely depends on monsoon for irrigation purpose. Its economy relies on farming of the nation. Due to the fact that its farming is based on natural monsoon, it is necessary for the farmers to know about when a rainfall occurs. In addition, rainfall information also helps people to become aware of any likely disaster which can cause damage to lives and property. This research carried the comparative study on the performance of Random Forest, Decision Tree, and Support Vector Machine classification algorithms for the prediction of rainfall on the dataset recorded by various weather stations of Nepal. The performance of these three classification algorithms were assessed using accuracy, precision, recall and F-measure. The obtained prediction result based on the performance measures showed that the Random Forest outperformed all the competing algorithms as it had highest accuracy value of 80.56 %, precision value of 74.50%, recall value of 76.50% and F- measure value of 75.50%.