Application Of Fuzzy Classification System Based On Genetic Algorithm In Stock Analysis

The stock market is not only an important part of a country’s economy, is also an important investment loved by the nationals.The stock market particularity influenced by national policies, investment and macroeconomic psychology, among other factors, investors appear to be in a very complex and unpredictable system. Because of the complex features of stock market uncertainty leading up to it particularly difficult to build model. During these years the development of the stock market, is growing larger and larger transactions, how to find useful information in these data become an problem to financial workers and majority of investors. Data mining technology is an important method of processing large data, it can obtain potentially valuable information from large amounts of data, so the use of the stock transactions in this predictive analysis techniques has theoretical and practical significance[1].In this paper, research and case study method of combining algorithm is proposed to obtain understandable fuzzy rules based on genetic algorithm method is applied to data classification, to achieve future data classification and prediction. The main work includes:First, according to the fuzzy classification model advantages in dealing with the issue of classification, genetic algorithm combined with the advantages of dealing with global optimization problem, and proposes a new fuzzy rule extraction method. Using a standard set of data on the UCI carried out experiments to prove its effectiveness.Secondly, to explore the data pre-processing properties of the discrete knowledge. Stock data for each attribute value is characterized by continuous proposes a fuzzy kmeans algorithm(FCM) for using the idea of fuzzy clustering, fuzzy partition on continuous attributes, building fuzzy subset, which will help build genetic algorithm coding and fuzzy classification system.Finally, the paper establishes a fuzzy classification model applied to data mining stocks, select the "SPDB" 483 historical transaction data indicators as the last two years of data collection, which in addition to "up down" for the category attribute, the remaining 30 It is characterized by a property. To achieve centralized extract fuzzy rules in the stock data and verify its ability to classify the data used to train 2/3, 1/3 is used for authentication. Follow the steps 1. Data preprocessing: Clear noise data, characteristics and classification of property attributes Pearson correlation analysis, selecting high correlation attribute as a characteristic property, streamline data sheet. 2. Property fuzzy division: first normalization of each attribute, and then use the FCM clustering algorithm fuzzy division. 3 using improved genetic algorithm fuzzy rules coding, crossover and mutation operations have evolved through high fitness individual fuzzy rules.