Different period time series forecasts integration as a tool of increasing the accuracy of stock return prediction

Abstract

One of the key factors in ensuring the efficiency of investment is ability to obtain high-accuracy stock return forecasts. Nowadays, there are many different prediction methods, ranging from simple moving averages to neural networks and other sophisticated methods. The analysis of forecasting methods and proposals to integrate separate forecasts to improve prediction precision was carried out in this article. The method of separate forecasts integration, based on forecasting accuracy in past periods, was proposed. The empirical study validating the suitability of proposed method in integration of fore-casts obtained analysing single and weighted moving averages (SMA and WMA) of different period time series was carried out.