| The stock market is such an important channel for financing, that every movement in the stock market is concerned by all kinds of news media. The news media offers authoritative, objective and true information that reflects not only the macro aspects of the stock market such as the politics, economics, law and so on, but also the micro aspects about the state of operation about a listed company. News media on one hand disseminates important information to the investors, on the other hand guides the public opinions. Along with the accelerating process of informationization, the news media shows a more and more important role in the dissemination of knowledge, troubleshooting and the supervision of public opinions, that impulses the development of the society. The news can affect the stock price, which has been universally recognized by the public. Previous studies have confirmed that the news has the ability to predict the stock prices, no matter it is in foreign or domestic market. The stock market, which contains numerous participants, is an extremely complex risky market, in which all the investors are very unequal in rationality and risk tolerance. Because any news report is likely to lead prompt and excessive reacts in the market, the investors must learn how to quickly master the information which can really affect the stock market in a news context to survive. The Internet has brought great timeliness and convenience to information transmission, so that internet news media has become a great challenge to the traditional papers. On one hand, many traditional papers gradually shift their focus to the Internet, on the other hand, more and more newly rising web media begin to release news reports on the Internet, So that the Internet financial news media become a great impetus in the development of the stock market.With the global Information explosion, the Internet financial news is bringing tremendous information, which grows in index way, into the stock market. According to Moore’s law and Metcalf s law, the web information not only can be consumed without loss, but also can generate new information during its spread, so the internet news will bring great external effects to the investors that may submerge one stockholder. The fast expansion of the information has brought great convenience to the investors as well as huge crisis. As the amount of information that one investor or one organization faces is far more than their tolerance, they had to cost lots of time and energy to process a mass of irrelevant and redundant information, so that they can avoid being misled by wrong information. Compared with traditional paper media, the competition of Internet media is more intense, that different Internet financial news media began to provide various promotion measures and comprehensive services one after another in order to attract investors. However, different Internet financial news medium are different in number of users, editing ability, service level and so on, which reflects that they will not have the same effect on the same stock market. Therefore, the past studies which take all the news, that comes from different sources, as one are not reasonable. We take this as a start point and did adequate researches on different news source.Facing the vastness of the Internet financial news, how can the investors effectively identify the most important information that is related to the investment? The Google search engine has established a scoring system, which is called PageRank, to rank the importance of every websites based on their link structures. Google sort the information from the most important website to the top of the search list, that the users can effectively obtain the most important information. Inspired by the search engine website ranking, we expect to establish a set of all Internet financial news website ranking mechanism, so that the news from the most influential website will has the most remarkable stock-price forecast effect. Once a news report is released, investors can quickly recognize whether the news is of importance and what effect it has on stock price, so they can make the right investment decisions in the first time. We are wondering if the PageRank value can be directly used to measure the financial news websites’effect on the stock market? Previous studies have confirmed that, the information of a news report can not be covered by its headline or its amount, the news text content has a considerable "soft strength" that can not be neglected. The kind of "soft strength" on one hand shows the fundamental information of a listed company, on the other hand guide the investors emotion so that to affect their decision-making. with the continuous development of financial intelligence technology, more and more scholars take the news text into account, and verified the influence of news context on the stock market by plenty of empirical studies. As we know, the PageRank algorithm used by Google search engine is a representative of the general social influence to all Internet users of a website, and this algorithm doesn’t take the news context into consideration. We believe that simply use the PageRank value to measure the influence of Internet financial news media in order to rank all the websites is not accurate, we need to seek one more scientific way.This paper mainly includes two research sectors:(1) Through the analysis of financial news text prediction ability of price, we discussed the influence of the financial news from Internet on stock market to a deep extent. In this sector, we studied how to build a forecasting model with the Internet news, how to choose the prediction time parameter, how to verify the validity of our model, and how to choose the news Quantitative method.(2) Construction of Internet financial news media ranking system. In this sector, we made full use of the prediction model built by in the fist sector to research the different influence of different websites. We fist separated all the Internet financial news according to their website sources, to discuss whether news from different resource will have different predictable ability on stock price. Then, we use the different predicable ability to build the indicator of website influence rank. At last, we compared our influence rank method to PageRank value, and explained the rationality of our ranking system.Our study belongs to the scope of financial intelligence, which is a combination of financial theory and information technology. Therefore, we use text processing techniques to analyze the internet news context. The techniques used in our study can be classified into two categories:one is used to collect the news, including web crawler and bloom filter; another is used for text mining, including FudanNLP, which is used for the quantification of news text, support vector machine(SVM), and Support Vector Regression(SVR),that is applied for build the stock price prediction model. First of all, through the use of web crawler and bloom filter, we captured all the Internet news text and stored the text information locally according to a prescribed manner. Next, we use FudanNLP Natural Language Processing to segment the news text and tag the part-of-speech of every word, so that we can extract the "key words" in a news text, namely "proper noun+emotional words". Third, we converted on text into a space vector, and applied the TF*IDF method to assign the vector by using SVM. Finally, we constructed a stock price predict model by employing SVR, so that we can discuss the different influence of various internet financial news media to the stock market.I have mainly used three kinds of data in this study:the high-frequency stock data, the Internet financial news and financial emotion lexicon. On the basis of previous scholars’conclusion——news has very short influence on stock price, and in the consideration of information timeliness, we use stock high-frequency data at the second level, which is a great breakthrough to the former studies. Meanwhile, the Internet financial news also provided the release time which is also at the second level. Then we can match the stock price and the Internet financial news accurately. The financial emotional lexicon we used in our study is on the basis of Loughran and Mcdonald lexicon. As the the Loughran and Mcdonald lexicon only contains English edition, we then translated it into Chinese and build a machine learning principle to expand it. The financial emotion lexicon we used is very significance to China’s Internet news market.There are five useful findings in our study:(1) The Internet financial news can indeed affect stock price. Just use the news text information and the current stock price, we can predict the actual future stock price.(2) The optimum prediction time window of Internet financial news to the stock price is26minutes. We can make a profit if we use the forecasted results to guide our trading.(3) The best way to quantify one news text is using "proper noun+emotion words"."Proper noun" reflects the fundamental information of a listed company, and "sentiment word" is used to describe the emotion hidden in the news. Only the combining of the two aspects can comprehensively represent one whole news context.(4) Different news from different web sources has different impact on stock price. Taking all the news as a whole will increase the prediction error.(5) PageRank value is not suitable for representing the influence ranking of an internet financial media, however, the indicator we constructed on the basis of stock price prediction decision accuracy of different websites has the rationality. |
PDF Full Text Request