FORECASTING MODELS WITH HYBRID TIME SERIES MODELING

The science of statistics is an essential part of each and every domain of human activity and is widely applied in framing policies and to formulate decisions in diversified areas covering socio economic, physical, Biological, engineering and social sciences.

Time series is a popularly using in modern era of various sectors. In Time Series analysis, theoretical and empirical findings have suggested that integrating various types of forecasting models can be an effective way to improve the predicting performance of each individual model. It is especially occurred when the models in the ensemble are quite different. Hybrid techniques that decompose a time series into its linear and non-linear components are one of the most important kinds of the hybrid models for the time series forecasting. Several researchers have focused and shown that these models can outperform single models.

An attempt is made in this book to forecast the daily prices of silver, gold metals and foreign exchange rates of Indian rupee (INR) against US dollar (USD) using conventional time series models, artificial neural networks (ANN) and Hybrid models and a comparative study is carried out to examine the forecasting capability of these models.

The present book is divided into six chapters including the introduction chapter and the chapter wise summary is given as follows: Chapter1 will introduce there search problem, objectives, fore casting methods, research data, research contribution and the significance of the book. In chapter 2, a brief review discussed on Box-Jenkins methodology, Feed Forward Neural Networks and Hybrid models.

Chapter 3 is a fabricated development of autoregressive integrated moving average (ARIMA), artificial neural networks (ANN) and Hybrid model for daily prices of silver in India. Forecasting the daily prices of silver in India using ARIMA, ANN and Hybrid models are estimated and compared. The building of ARIMA, ANN and Hybrid models for daily prices of gold in India is comprised in chapter 4. Chapter 5 comprises development of ARIMA, ANN and Hybrid model for foreign exchange rates of INR/USD is described in detail. The performance of models is evaluated by using error measures and forecasts are also estimated and results are shown that Hybrid model performing well at fitting stage as well as forecasting stage for forecasting foreign exchange rates than that of ARIMA and FFNN models. The MGN and Boot strap tests are applied to test the equal prediction accuracy of the fore casting models. Chapter 6 is presented with the final results, discussions, and comparisons. With above all, a researcher can adopt and also search for future recommendations.

We would like to take this opportunity of expressing our sincere gratitude to the various research organizations, colleagues, Research scholars, friends and family members for their constant support and encouragement. We have a strong belief that there is a scope for further improvements. Although proper care has been taken to avoid all kinds of errors, we shall be grateful to the teachers, research scholars and readers for their valuable suggestions and advices.