Real Time Rainfall Prediction for Hyderabad Region using Machine learning Approach

A novel technique for the Prediction of Monsoon Rainfall using Hybrid Machine Learning Algorithms and their ensembles, operated on Real time Hyderabad Rainfall data is Proposed. Over the years many models have been developed for the prediction of rainfall, An Accurate Prediction of Monsoon Rainfall for farmers is necessary for centuries together. As regression algorithms fail to generalize the outcomes with the mean of rainfall being near 0(no rainfall). We have come up with a Hybrid Technique that uses the classification of rainfall into Binary Data, there upon which applies the Regression Algorithm, which significantly improved the result. Out of the many tests conducted on various algorithms, the Hybrid Model of Random Forest Classifier with Multi-Layer Perceptron Regressor has given the best outcome with a practical Accuracy of 0.92537 and with a Mean Absolute error of 1.063540 tested over a period of 3 months. Furthermore, we have elaborated different Hybrid Model compositions used in ensemble learning for the improvement of small scale outputs. These results are tested on Real time data obtained on Hyderabad rainfall data and compared with real time.