Research On Stock Price Prediction And Stock Selection Strategy Based On LSTM Model

The overall wealth of China’s banking industry ranks first globally,while the size of the stock market,bond market,and insurance market ranks second globally.At the CPC’s 20 th National Congress in October 2022 it was declared that the financial system must be restructured,financial services must be more beneficial to the real economy,direct financing in social finance should be increased,and the multi-stage capital market should be further advanced.To further advance the reform and growth of the capital market,strategic plans should be made.At the December 2018 Central Economic Conference,it was proposed that the capital market,a major factor in financial business,should be further reformulated to form a standardized,transparent,open,dynamic,and resilient Chinese capital market.This has become a major focus in the structural reform of the financial supply side,with deepening the reform of China’s capital market being a top priority." The positioning and strategic importance of the new era capital market are thus clear.The securities market has seen a surge in the use of financial analysis software due to the swift advancement of computer technology and big data analysis technology.This has provided a significant technological basis for capital market investment,with the stock market boasting a large number of participants and being highly active.The irrational speculative tendencies of individual investors,which lead to profit-seeking and selling,are not conducive to the healthy growth of the stock market.However,if technology can be employed to forecast stock prices and offer investors investment advice,it will foster the capital market’s healthy development and enhance investor returns,a concept with both theoretical and practical implications.Gradually,the emergence of computer and big data technology has enabled machine learning to become a part of people’s day-to-day lives,and the amalgamation of machine learning and quantitative tactics has drawn a great deal of scientific interest.At present,the most common machine learning algorithm is the stock market prediction monitoring learning algorithm,such as decision tree,random forest,neural network and support vector machine.The quantitative strategies for predicting stock prices based on these models have performed well in terms of returns.A sliding window rolling method was employed to generate data samples for training the neural network model,which was then pre-processed on the daily frequency raw trading data of the Shanghai Stock Exchange 50 Index constituent stocks from 2011 to 2020.This article introduces a deep learning neural network algorithm for predicting stock price trends,utilizing the neural network LSTM model to forecast stock price fluctuations.By contrasting the model’s prognostication accuracy with various input variables,the input variables of the model examined in this article were established and the fundamental trading indicators of individual stocks were eventually acquired.A total of 20 distinguishing elements,such as the Shanghai Stock Exchange 50 index indicators and individual stock technical indicators,were taken into account.Second,various control experimental groups were formed,and by altering the internal parameter combinations of the model,the most advantageous parameter combination for forecasting precision was chosen while investigating the effect of different parameters on the model’s prediction capability.A comparison of the LSTM model’s prediction performance under this parameter combination to that of the BP model,the model,and the LSTM model was made.Subsequently,the backtesting period’s market status was divided into bull and bear markets,and a quantitative stock selection strategy was formulated based on the rise and fall probabilities predicted by the LSTM model as stock selection conditions.Backtesting was utilized to assess the strategy’s efficacy in both a bull and bear market state.This empirical study shows that:(1)The LSTM prediction model performs well in predicting the rise and fall of the Shanghai 50 stock index constituent stocks.The LSTM model,in comparison to other neural network based models,is superior in forecasting stock prices.(2)The quantitative stock selection strategy based on the LSTM model can yield significantly higher returns than the benchmark in various market conditions,implying that the model’s output fluctuation probability can be utilized as a stock selection factor to acquire extra returns.Quantitative stock selection can be greatly facilitated by the fluctuation probability,providing investors with a benchmark for their investment when making their selection.