Study On Trend Analysis Of The Financial Field Based On Text Mining

A large amount of data information has been accumulated in the field of finance (including a variety of financial markets). The data include both structured numeric data and large amounts of unstructured text data. Investors can make use of related technologies to dig out useful information in massive data from financial markets, and then to guide their investment behavior and help make right decisions. In this study, I investigate the impact of instant Chinese financial news reports on price trend of financial markets (mainly the Chinese stock market) with text mining techniques.Studies suggest that the impact of the financial news on the stock price is basically a classification problem about financial news; the news can be divided into positive and negative categories. Positive news will promote the stock prices rise, the negative news will cause the price fall. In this study, I mainly focus on Chinese financial news text, and investigate text classification techniques in the field of text mining.In this thesis, text mining technologies are briefly introduced first. Then text classification techniques are studied in detail. Text preprocessing, text representation and text classification procedures were deeply investigated. Based on the theory of text categorization, a financial news classification system is designed and implemented. In text preprocessing module, the number of words of each news report is reduced by retaining nouns and verbs, and removing stop words. In news representation module, four feature selection algorithms are implemented. In text classification module, news reports are used for training and testing by Libsvm software. Finally, the classification results are presented, and the corresponding judgments on stock price trend are given.News reports for experiments are selected from the finance news channel of Sina. These documents in a certain period are about some companies which belong to banking, non-ferrous metals, or coal and oil industry. The implemented financial news classification system is used to classify test news from the three industries. The changing direction of the corresponding short-term stock price trends are gotten according to the classification results, compared the company’s actual stock price trends to get the stock price prediction accuracy, and the results are analyzed in detail.The experimental analysis shows that the stock prices of companies form non-ferrous metals and coal and oil industry are more vulnerable to the impact of related financial news reports.