Intelligent weather monitoring systems using connectionist models

This paper presents a comparative study of different neural network models for forecasting the weather of Vancouver, British Columbia, Canada. For developing the models, we used one year’s data comprising of daily maximum and minimum temperature, and wind-speed. We used Multi-Layered Perceptron (MLP) and an Elman Recurrent Neural Network (ERNN), which were trained using the one-step-secant and LevenbergMarquardt algorithms. To ensure the effectiveness of neurocomputing techniques, we also tested the different connectionist models using a different training and test data set. Our goal is to develop an accurate and reliable predictive model for weather analysis. Radial Basis Function Network (RBFN) exhibits a good universal approximation capability and high learning convergence rate of weights in the hidden and output layers. Experimental results obtained have shown RBFN produced the most accurate forecast model as compared to ERNN and MLP networks.