JISE

Financial time series are often considered to be difficult to model and unlikely to predict. In this study, we assume that financial time series are based on a stochastic series generated by a Markov decision process. Based on this assumption, we investigate two problems related to the identification of the price time series of financial instruments. We try to distinguish the real price-volume time series from the artificial one. First, we investigate whether there is any machine learning model that can distinguish between real price-volume time series and those with time horizon reversed. Then, we investigate whether there is any machine learning model that can distinguish the price-volume time series from the real one when they are subjected to random manipulations of different proportions. The data we use are the daily prices and trading volumes of six U.S. stocks and one crypto-currency BTC/USD. We apply Long-Short Term Memory (LSTM) as the main machine learning model for the binary classification due to its success in fitting time series data. Based on the experimental results, we give positive answers to the above two questions. Our results also partially support the conjecture that the dynamics of a financial time series are driven by an underlying Markov decision processes.