FRANN: Fuzzy Rough Set Based Artificial Neural Network for Prediction of Rainfall Status

Rainfall status prediction is a process of forecasting the future status or condition of rainfall for a particular region. Prediction of rainfall status is essential because heavy or low rainfall may have many adverse effects on agriculture, transportation, environment, living conditions and economic strength. So, accurate prediction of rainfall status is helpful in designing better action plans for water resource management, disaster risk reduction, sustainable development, emergency management and improving living conditions. Computational intelligence and approximate reasoning of conventional machine learning models are limited by incomplete or imperfect knowledge. However, the soft computing techniques have the capabilities to handle uncertainty, imprecision, partial truth and approximate reasoning to extract information in a number of ambiguous situations. In this aspect, a fuzzy rough set (FRST) based artificial neural network (ANN) model is proposed to predict the future rainfall status. Incorporating FRST method in ANN model can improve the prediction accuracy and reduce the overheads associated with both memory as well as computation time by selecting the dominant attributes and handling inconsistent data present in the information system. Rainfall status prediction accuracy of the proposed hybrid model is compared with two other hybrid models namely FRST based decision tree (DCST), FRST based support vector machine (SVM) and three other existing classification based prediction models namely DCST, SVM and ANN. The findings of the result suggest that the proposed model outperform its counterparts.