Predicting Time Series Using Integration of Moving Average and Support Vector Regression
Chihli Hung, Chih-Neng Hung, Szu‐Yin Lin · International Journal of Machine Learning and Computing · 2014
Time series prediction is one of the major tasks in the field of data mining.The approaches of time series prediction can be divided into statistical techniques and computational intelligence techniques.Most researchers use one specific approach and compare the performance with other approaches.This paper proposes a novel hybrid approach, which integrates traditional moving average models with support vector regression for the prediction of ATM withdrawals in England.The use of moving average modeling is not only for the purpose of smoothing but also for time series prediction.We treat a weekly median moving average as the benchmark.Based on experimental results, our proposed approach consistently outperforms the benchmark.