Stock Trading by Modelling Price Trend with Dynamic

Stock Trading by Modelling Price Trend with Dynamic Bayesian Networks Jangmin O 1 ,JaeWonLee 2 , Sung-Bae Park 1 , and Byoung-Tak Zhang 1 1 School of Computer Science and Engineering, Seoul National University San 56-1, Shillim-dong, Kwanak-gu, Seoul, Korea 151-744 {jmoh,sbpark,btzhang}@bi.snu.ac.kr 2 School of Computer Science and Engineering, Sungshin Women’s University, Dongsun-dong, Sungbuk-gu, Seoul, Korea 136-742 jwlee@cs.sungshin.ac.kr Abstract. We study a stock trading method based on dynamic bayesian networks to model the dynamics of the trend of stock prices. We design a three level hierarchical hidden Markov model (HHMM). There are five states describing the trend in first level. Second and third levels are abstract and concrete hidden Markov models to produce the observed patterns. To train the HHMM,

we adapt a semi-supervised learning so that the trend states of first layer is manually labelled. The inferred probability distribution of first level are used as an indicator for the trading signal, which is more natural and reasonable than technical in- dicators. Experimental results on representative 20 companies of Korean stock market show that the proposed HHMM outperforms a technical indicator in trading performances. 1 Introduction Stock market is a core of capitalism where people invest some of their asset in stocks and companies might raise their business funds from stock market. Since the number of investors is increasing everyday in this century, the intelligent de- cision support systems aiding them to trade are keenly needed. But attempts on modelling or predicting the stock market have not been successful in consistently beating the market. This is the famous Efficient Market Hypothesis (EMH) say- ing that the future prices are unpredictable since all the information available is already reflected on the history of past prices [4]. However, if we step back from consistently, we can find several empirical results saying that the market might be somewhat predictable [1]. Many of technical indicators such as moving averages have been developed by researchers in economic area [3]. There are some weak points in technical in- dicators. For example, if we use RSI, we must specify its parameters. The curves of RSI are heavily influenced by the parameters. Also, there are some vagueness in their interpretations, which might be varied according to the subjectiveness of the interpreters. In this paper, we propose a trend predictor of stock prices that can produce the probability distribution of trend states under the dynamics of trend and Z.R. Yang et al. (Eds.): IDEAL 2004, LNCS 3177, pp. 794–799, 2004. c© Springer-Verlag Berlin Heidelberg 2004 Stock Trading by Modelling Price Trend with Dynamic Bayesian Networks 795 price of a stock. To model the dynamics, we design a hierarchical hidden Markov model, a variant of dynamic bayesian networks (DBN). Given observed series of prices, a DBN can probabilistically inference hidden states from past to current. Also we can sample or predict the future from learned dynamics. To use an indicator of bid and ask signals, it is more natural to use our HHMM model than technical indicators. The resulting trading performance is compared with the performances of a technical indicator through a simulated trading on Korean stock market. 2 HHMM...

Website: eecs.harvard.edu | Filesize: 202kb
No of Page(s): 6
Download LNCS 3177 - Stock Trading by Modelling Price Trend with Dynamic ....pdf