Stock price prediction using fuzzy rule based systems

From a technical point of view, by predicting the stock price and making smart decisions in order to buy and sell on time, you can achieve a significant profit. This issue has led many researchers and analysts to predict stock prices by providing different methods. Since the nature of the stock market is very complex and fluctuating, it has been observed that the traditional and statistical methods that analyze the model assuming linear relationships are not very efficient. In recent years, it has been observed that methods based on artificial intelligence and data mining, which have the ability to solve complex models, have provided better performance. In this regard, in this research, the method of fuzzy rule-based systems is proposed for its high ability to model complex problems and also high interpretability for humans. Also, in order to homogenize the data of the problem, the density peak clustering method has been used in this research. RIPPER algorithm is used as one of the best rules extraction methods. These rules, which were initially created in crisp form, are converted into fuzzy rules in Mamdani fuzzy system. Finally, in order to optimize the created fuzzy system, genetic algorithm has been used. The measurement of the model error has been done using the Mean Absolute Percentage Error (MAPE) index, and the results show that the proposed model performs well compared to other similar studies

Keywords:

Prediction models , stock price , fuzzy logic , rule extraction , machine learning , clustering