Predicting the conditions of stock market using hidden Markov models

• Main Authors: Setiawaty, Berlian, Lestari, Amanah Yuni, Lesmana, Donny Citra, Ardana, N. K. Kutha
• Format: Conference Proceeding
• Language: English
• Subjects: Algorithms; Decoding; Markov chains; Securities markets; Stock exchanges; Viterbi algorithm detectors
• ISSN: 0094-243X; 1551-7616
• DOI: 10.1063/5.0230595

Stock market conditions constantly change, which is indicated by the movement of the closing values of a stock index, either increasing or decreasing. Suppose the stock market conditions are assumed not to be observed directly and form a Markov chain. In that case, the pair of stock market conditions and the closing values of a stock index can be modeled as a discrete hidden Markov model. Three fundamental problems in the hidden Markov model are evaluation, decoding, and learning problems. These problems can be solved successively using the forward-backward, Viterbi, and Baum-Welch algorithms. The Viterbi algorithm can be used to predict market conditions that are hidden behind stock index data. In this research, we predict the stock market conditions of the FTSE 100 index from 2018 to 2020. The results from the Viterbi algorithm show that there are three market conditions of the FTSE 100 index. These conditions can later be recognized as bullish, sideways, or bearish. A bullish market is characterized by an increasing trend in the stock index, while the tendency of a downward trend in the long term is called a bearish market. A sideways market is shown by price changes, either in the form of increases or decreases, at narrow intervals over the long term.