Research And Implementation Of Stock Price Prediction Based On The Analysis Of Sentiment In Text

Stock is becoming one of the major assets allocation ways among Chinese residents. But if lack of precise tendency forecasts or miss the appropriate trading time, the fickle price of stock will undoubtedly lead to profits shrinkage, even cause investment risks. Existing analysis methods of stock mainly depend on technology indexes of stock software, which is based on experience and artificial judgements. There are two problems in Such method: firstly technology indexes offers such a wide variety that different indexes have different application range. And the prediction of single index is unsatisfying, so combining different indexes to make decisions is necessary, apparently there are huge cost of study to investors in such way. Secondly, technology index itself has evolved from the objective facts which predicted in stock market through price, turnover etc. which cannot reflect subjective attitude and emotion of investors.To solve the problems mentioned above, this paper research on price prediction of stock, providing three invest signals(buy, hold, sell) to help the investors make decisions, mainly conducted from following two aspects.Predict the price of stock according to technology indexes. Firstly, define the goal of stock price prediction, and confirm the thresholds of buying, holding and selling respectively. The result of three invest signals are considered as predicted output. Then, quantize technology indexes such as MACD, RSI, KDJ, which are predicted input. Finally, choose the best parameters to formulate a stock-price prediction model which is based on BP neural networks.Predict the price of stock according to emotions of investors. In the first place, collect comment information of stock on guba website through web crawlers. Then, structure a specialized financial dictionary, Use N-gram algorithm to construct the word segmentation dictionary, Use PMI algorithm to extend emotions dictionary.Based on the contrast of tow sentiment classification methods,I Build a best-performed classifier.Finally,I compute the Investor sentiment index with the formly build classifier as an input of stock price prediction model.This paper exploratory investor sentiment indicators as variables to share price forecast model, the experimental results show that investor sentiment index has guiding significance for predicting stock price movements, can to a certain extent, improve the accuracy of the stock price forecast. And ultimately the integration of technical indicators and investor sentiment index, I formulate a well performed stock-price prediction model which is based on BP neural networks.