Stock Trading Signal Prediction Using a Combination of K-Means Clustering and Colored Petri Nets (Case Study: Tehran Stock Exchange)

Stock markets are attractive in nature for investors to gain profit. However decision making about suitable points of trading is a challenging issue, due to various properties of stocks, unstable values and data frequencies. Predicting stock price movements and discovering turning points using technical indicators, for the sake of data frequency reduction in short-term, is a preferred choice in comparison with price forecasting which commonly uses fundamental analysis. In this ambit, this paper proposes a Colored Petri Net model combined with k-means clustering decision making rules to predict stock trading signal, namely buy, sell, and hold, enhanced by a strength coefficient in a 7-step process. The paper focuses on Tehran stock exchange as case study in a two-year time interval. Simulation results implies superiority of proposed model against other state-of-the-art approaches, i.e. artificial neural networks, decision tree, and linear regression, with the accuracy rate of 88% in term of correctly classifying.