Time series prediction of retirement mutual fund values using optimal window size selection and support vector regression

Retirement Mutual Fund (RMF) is an alternative way of saving for retirement. Most of the RMFs provide a moderate growth of profits. However, there are several investment plans being offered such as low-risk plans and high-risk plans. An investor has to make a decision on what to buy based on the past records and several factors. One of the main factors is to study the trend of fund values in the future. This paper proposes to use a heuristic technique to select the optimal window size of data needed for accurate prediction and to use Support Vector Regression (SVR) to predict the 1-day ahead values of RMF based on time series data from the past. The proposed algorithm was evaluated on 91 RMF plans from eight risk levels. The statistical results reveal that most of the low-risk plans use more time series data in the past to predict the future values than the high-risk plans. Furthermore, by using the optimal window sizes, the overall accuracy rate of the proposed algorithm is more than 89.00%.