Based On The Integrated Study Of Stock Market Forecast And Trading Strategy Research

With the development of national information construction,applying advanced computer technology to the financial market,learning the rules of trading from massive data,and formulating strategies to guide trading has become a vast range of investment methods.More and more people are participating in stock investment after people ’s material life dramatically improved.Procedualized stock price research can not only avoid the subjective misjudgment caused by investors’ emotional changes,but also dig out hidden rules that can not be found by human beings from huge data,so it has attracted the attention of many experts and scholars.For the deficiency of manual analysis law of stock,investors usually want to get one of the most suitable for the current stage of the intelligent investment strategy,meanwhile obtain higher revenues with a low risk.This article aims to combine the stock market results predicted by the machine learning model with trading strategies to expect that the formation of new trading strategies can get higher yields,lower transaction risk.Firstly,build a stock price prediction model based on machine learning.The XGBoost model,Light GBM model,and Cat Boost model in binary classification problems in recent years,are used to predict the rise and fall of stock prices.On this basis,the voting based method is selected to conduct model fusion to form the fusion model for predicting the rise and fall of stock prices.The input data of the model is composed of two parts.The first part is the stock trading data and technical indicators,and the second part is the micro-blog text data.Since the technical indicator of the stock is calculated by the trading data of the stock through different mathematical methods,there is a specific correlation between the two.So in this paper,the principal component analysis is used to process the data.The second part of the data is obtained by crawling the micro-blog website.The crawled micro-blog text data is used for sentiment analysis,and the generated sentiment values are averaged by day.Connect the above two parts of data by date to form the data set.The basic model is built based on the three integrated learning algorithms i.e.,XGBoost,Light GBM,and Cat Boost,while the fusion model is constructed by the voting method.Secondly,build the trading strategy based on the predicted results,which is divided into two steps.The first step is to use a traditional trading indicator to build a basic trading strategy model to find the best trading strategy suitable for 2018 to 2020.In this study,the interval rolling training method finding the best trading strategy combination in the validation set interval is used.After that,transform the rise and fall results predicted by the model into trading signals in a certain way,it then guides trading operation jointly with trading signals from the trading strategy portfolio.Finally,this paper combines the prediction results of stock price rise and fall with the basic trading strategy to form a new trading strategy applied to the verification set interval to conduct simulated trading.The trading results show that the trading strategy based on the forecasting model has a higher profit rate than the traditional trading strategy alone.The maximum profit rate can reach 3.2 times that of the basic strategy,and the maximum retracement rate is reduced by 5.4%.Therefore,it can be concluded that the trading strategy based on the prediction results of the fusion model can obtain higher return rate than the basic trading strategy,which provides a new way for investors to analyze the market situation.