MODELLING AND EVALUATION PERFORMANCES WITH NEURAL NETWORK USING CLIMATIC TIME SERIES DATA

Neural network model is an alternative powerful data modeling tool capable of capturing and representing non-linear relationships between variables. The methodology used in this research article were Artificial Neural network (ANN), Autoregressive Integrated Moving Average (ARIMA) and Fuzzy Time Series (FTS) models. The study revealed that, since ANN forecast performance evaluation has the lowest value of sum of square error (SSE), Mean square error (MSE) and Root mean square error (RMSE), then ANN model outperforms ARIMA and FTS models. The ANNmodel was found to be more efficient with minimum parameters and capable of handling the non-linearity that characterized climatic data series.