Authors : Zohreh Sadrnezhad, Majid Vafaei Jahan and Sina Zarrabi Darban
Abstract: In this study, a new concept named Markovian-distance is presented. Markovian-distance is defined based on Markov-Chain concepts but in contrast with Markov-Chain previous states have an effective impact not the prior state alone but also depends on all former states. This implication has various usages such as extracting similar patterns. In this paper case study of predicting stock trends behavior using Markov-distance is described. This approach is able of extracting similar patterns based on resembling transfers between different states of stock price (increase, decrease and price stability) and similar patterns with differentiated lengths. To predict stock trends behavior in the proposed method, we extracted similar patterns on Relative Strength Index (RSI) indicator because in this indicator speed and rate of price changes are measured over a time period and it also has different levels in which every one of them has a distinct conduct. In addition, every similar pattern in every level will probably follow a specific trend in future. So by using this indicator among similar patterns, patterns which are at a resembling level to the RSI current volatility level could be extracted to have a more accurate prediction. Finally, the evaluation of the proposed method is done by using Isfahan Steel Stock on Tehran stock market and Apple Corporation Stock dataset. Results show that the proposed method predicts the price trend in Isfahan Steel Stock with 88.5% and Apple Stock with 84.87% accuracy.
Zohreh Sadrnezhad, Majid Vafaei Jahan and Sina Zarrabi Darban, 2020. Extracting Similar Patterns to Markovian Distance: A Case Study on Predicting Stock Trends Behavior. International Journal of Soft Computing, 15: 41-43.