Prediction Of Baijiu Stock Based On Machine Learning Algorithm Model

With the development of China's financial market,stock price forecasting has been paid more attention by people of various circles.Stock forecasting can bring huge profits to investors in the financial market,and can also become a tool of risk management.With the development of machine learning and artificial intelligence technology,it has achieved amazing results in many fields.Machine learning and deep learning algorithm model can process massive data and quickly process the analysis rules.Many scholars try to apply a variety of machine learning models in the financial field,and the quantitative investment analysis technology based on mathematical model has been widely used.Stock data has the characteristics of non-linear,high noise and non-stationary.It is of practical and theoretical significance to find a reasonable way to comb stock data,build a non-linear machine learning model,and explore the law of stock price rise and fall.Nowadays,liquor sector stocks are highly concerned.This article mainly predicts and studies the price rise and fall of liquor stock market.It selects all liquor stock data in the Shenwan industry classification for research.First,considering the financial market has many disturbance factors,stock time series data It contains a lot of noise,and the data is preprocessed using wavelet threshold denoising to extract useful signals from stock data.The empirical part takes two stocks as an example to show the comparison chart before and after the closing price of the stock is reduced by wavelet noise.This article mainly uses LSTM,support vector machine and XGBoost model to predict the stock price of liquor.The LSTM long short-term memory network can mine time information and can fit the forecast stock price well.XGBoost is an emerging Boosting integrated algorithm model,which is efficient and accurate.,Support vector machine algorithm is simple in principle and has good promotion ability.This article selects stock data from March 2,2015 to March 2,2020 as a sample.The data comes from the wind database.The data processing and modeling process is completed in the python language,using market indicators and related technical indicators,and through features Engineering augmentation constructs more features as input variables of the model.The empirical results show that the three machine learning models all have excellent predictive and fitting capabilities for the trend of stock closing prices.This article uses MAE,RMSE,MAPE,R2 to evaluate the prediction effect of the model.LSTM performs the best.It also proves the feasibility of the emerging machine learning algorithm XGBoost in stock price prediction.Compared with other algorithms LSTM,support vector machine,etc.,XGBoost runs faster,has certain advantages.Based on the original algorithm model,this paper continues to try to optimize the model,uses genetic algorithm to apply to the XGBoost parameter tuning process,uses the genetic algorithm's good optimization ability to select a better parameter set,and iteratively improves the model prediction effect;The encoder-decoder encoding-decoding structure is added to the LSTM model,and the attention mechanism is used to try to improve the model's predictive ability.