Investment Funds Management Strategy Based on Polynomial Regression in Machine Learning

摘要

This paper presents the results of an investment strategy simulation. The strategy is based on common regression models in a time series, which yields the decision. A simple polynomial regression was the basic method used to achieve short-term value forecasts in the time series. Base params (number of steps in the past and a degree of a polynomial) were set based on a machine learning algorithm. The strategy is improved with some additional original (constitutes by the authors) parameters because using only the regression proved to be completely ineffective. Financial markets with bidirectional transactions (long and short transactions), as well as only long transaction markets, were both taken under research.