Stock Price Trend Forecast Based On Securities Analysts' Reports

According to the EMH(Efficient Markets Hypothesis)and Behavioral Finance,this paper assumed that the securities analysts' reports are related to the trend of stock market,and realized the stock price forecast by analyzing the securities analysts' reports from the view of buyers.In this paper,we collected the securities analysts' reports released between April 2017 to August 8755 as the research samples,and used natural language processing,document sentiment analysis and supervised learning technologies.And analyze the confidence of the securities analysts' reports.Securities analysts' reports are first segmented and then converted into data that a computer can understand and learn through the natural language processing technologies.This paper uses the TF method to extract textual features,uses the TF*IDF algorithm to carry out the feature weighting,and constructs a characteristic dictionary with 700 dimensions.According to existing research results,we build an emotional dictionary with respect to securities analysts' reports,where the weights of emotional words are determined by comparing with the positive and negative emotional words.Furthermore,some especial emotional words are quadratic weighted.Based on emotion dictionary,this paper makes an emotional analysis on the securities analysts' reports.And classify the samples according to the actual rise and fall of the stock.We transform samples into text vectors composed of feature weights and use the SVM(Support Vector Machine)to learn a predictive model.During building the prediction model,we test the linear kernel function and the RBF(Radial Basis Function)respectively to train the model with a 10-fold cross validation.The grid search optimization method is used to pick up the best parameter.Relevant experiments were completed,the experimental process and data were recorded,and the experimental results were obtained.Finally,the efficiency of the system was analyzed.The study found that the confidence of the securities analysts' reports is insufficient,and there is a clear positive emotion bias,which is not strongly correlated with the actual fluctuation of stock.The emotional dictionary and the emotion analysis method proposed in this paper are suitable for the emotional analysis of securities analysts' reports.The limitation of this paper is that the distribution of report categories is uneven.Largerscale and more balanced sample gatherings will be more conducive to the study of this paper.