期刊名称：International Journal of Computer Science and Engineering
印刷版ISSN：2278-9960
电子版ISSN：2278-9979
出版年度：2021
卷号：8
期号：1
页码：30-38
DOI：10.14445/23488387/IJCSE-V8I1P106
出版社：IASET Journals
摘要：Weather forecasting is important for the daily life plan of the person, but the agriculture sector is also dependent on the weather condition and several industries. This research work reflects the prediction of the metrological parameter in the time series, i.e., prediction of Rainfall by using ANN (Artificial Neural Network) based model NARX (Nonlinear Autoregressive with exogenous input). In this research paper, more than a few ANN models that rely on real-time sequence recurring NARX-based ANN techniques are initiated, trained, and tested with different parameter settings to find the network model's best possible output to its most-wanted prediction function. The model network's performance is evaluated on the basis of the Mean Squared Error (MSE) performance of the model when the data sets are trained, validated, and tested. Although one step ahead of prediction, multi-phase ahead prediction is more complicated and difficult to carry out because of its underlying added complication. Therefore, the findings found in this work provide useful and helpful suggestions for the NARX-ANN model parameters, particularly the choice of hidden layer size and self-regressive leg terms for an efficient predictor of multi-step time series. This study aims to build and train a network that can predict and predict the weather portion of precipitation by optimizing the parameters of the neural network.
关键词：rtificial Neural network (ANN); NARX model; Outlier recognition; Time Series forecasting; Rainfall prediction.
