Prediction of Financial Performance Using Genetic Algorithm and Associative Rule Mining

The proposed system introduces a new genetic algorithm for prediction of financial performance with input data sets from a financial domain. The goal is to produce a GA-based methodology for prediction of stock market performance along with an associative classifier from numerical data. This work restricts the numerical data to stock trading data. Stock trading data contains the quotes of stock market. From this information, many technical indicators can be extracted, and by investigating the relations between these indicators trading signals can discovered. Genetic algorithm is being used to generate all the optimized relations among the technical indicator and its value. Along with genetic algorithm association rule mining algorithm is used for generation of association rules among the various Technical Indicators. Associative rules are generated whose left side contains a set of trading signals, expressed by relations among the technical indicators, and whose right side indicates whether there is a positive ,negative or no change. The rules are being further given to the classification process which will be able to classify the new data making use of the previously generated rules. The proposed idea in the paper is to offer an efficient genetic algorithm in combination with the association rule mining algorithm which predicts stock market performance. Keywords— Genetic Algorithm, Associative Rule Mining, Technical Indicators, Associative rules, Stock Market, Numerical Data, Rules INTRODUCTION Over the last decades, there has been much research interests directed at understanding and predicting future. Among them, to forecast price movements in stock markets is a major challenge confronting investors, speculator and businesses. How to make a right decision in stock trading extracts many attentions from many financial and technical fields. Many technologies such as evolutionary optimization methods have been studied to help people find better way to earn more profit from the stock market. And the data mining method shows its power to improve the accuracy of stock movement prediction, with which more profit can be obtained with less risk. Applications of data mining techniques for stock investment include clustering, decision tree etc. Moreover, researches on stock market discover trading signals and timings from financial data. Because of the numerical attributes used, data mining techniques, such as decision tree, have weaker capabilities to handle this kind of numerical data and there are infinitely many possible ways to enumerate relations among data. Stock prices depend on various factors, the important ones being the market sentiment, performance of the industry, earning results and projected earnings, takeover or merger, introduction of a new product or introduction of an existing product into new markets, share buy-back, announcements of dividends/bonuses, addition or removal from the index and such other factors leading to a positive or negative impact on the share price and the associated volumes. Apart from the basic technical and fundamental analysis techniques used in stock market analysis and prediction, soft computing methods based on Association Rule Mining, fuzzy logic, neural networks, genetic algorithms etc. are increasingly finding their place in understanding and predicting the financial markets. Genetic algorithm has a great capability to discover good solutions rapidly for difficult high dimensional problems. The genetic algorithm has good capability to deal with numerical data and relations between numerical data. Genetic algorithms have emerged as a powerful general purpose search and optimization technique and have found applications in widespread areas. Associative classification, one of the most important tasks in data mining and knowledge discovery, builds a classification system based on associative classification rules. Association rules are learned and extracted from the available training dataset and the most suitable rules are selected to build an associative classification model. Association rule discovery has been used with great success in International Journal of Engineering Research and General Science Volume 3, Issue 1, January-February, 2015 ISSN 2091-273