Development of Back-Propagation Neural Network for Prediction of Chaotic Data Time Series. A case study of Indian monsoon rainfall over a smaller geographical region
Long-range monsoon rainfall data time series over a smaller geographical region like `districts/subdivision' is representing chaotic in nature. It is found that, identification of internal dynamics of its time series as well future prediction is exceptionally complicated and for all time it is worldly nervousness. At present, it is a vital challenging task for meteorological services all over the world. It is found that, ANN technology has produced sufficient skill especially in prediction. Thus four separate, back-propagation neural network (BPN) models have been developed and verified. These models are:
BPN model in deterministic forecast
BPN model in parametric forecast
Principal Components BPN model
Hybrid BPN model
The model in deterministic forecast has produced excellent results, explained by strong relation between dependent (i.e., current year rainfall), and independent variables (i.e., past recorded rainfall data time series). Likewise, other models has also produced excellent results, explained by strong relation between their dependent (current year rainfall), and their corresponding independent variables (i.e., predictors). These models have been implemented through Java programming language. The entire development process and its observations are discussed in this book and expected that these observations will be extremely helpful for Scientists, Engineers, Academicians, and Research scholars who actually desire to exercise BPN in their chosen field of applications.