Research And Implementation Of Price Prediction Model Based On Machine Learning

With the continuous economic and social development,commodity price has the unique function as a lever to regulate the economy.Forecasting price is an important component of making macro-level decision and micro-management,which is of great significance to people's life.Because of the many factors that affect the price of commodity,price forecasting has become the difficulty of research.There are many commodity price forecasting methods,such as forecasting price trend based on statistical correlation,price forecasting model based on simple time series;intelligent price forecasting method based on machine learning and deep learning.The method based on statistical correlation is mainly based on empirical prediction and has certain subjectivity and limitations.The price forecasting method based on simple time series improves the accuracy of forecasting,but because this kind of method only considers the price data within a short period of time,the utilization of historical data is insufficient,which leads to the unsatisfactory result.With the development of machine learning and artificial intelligence,some advanced integration algorithms and deep learning prediction methods based on time series have high accuracy and robustness For non-linear and time-series data,these algorithms gradually become an inevitable choice to solve the problem of price forecasting.Based on the in-depth analysis of emerging machine learning algorithms such as LightGBM and Convolutional Neural Network(CNN),a price forecasting algorithm for CNN-LightGBM combination is proposed.The features of data were automatically extracted from CNN,and the extracted features were set into LightGBM for supervised training,finally the CNN-LightGBM combined price forecasting model was obtained.The model has a significant improvement in accuracy compared to the single use of CNN or LightGBM.Aiming at the fact that price data is also affected by other factors besides time series,a comprehensive multi-factor-influenced LSTM price forecasting method based on LSTM deep learning algorithm is proposed.This method not only utilizes the memory of lstm to historical data,but also introduces the influence of external factors on price through the whole connection layer,which provides a new way to solve the problem of price prediction.Compared with BP neural network,the results show that the method has higher precision and better stability.This paper divided the actual vegetable and fruit price data into training set and test set,and the two proposed price forecasting algorithms were used to train or test the data set,the results achieved the expected goal,which showed that the proposed price forecasting algorithm is feasible and popularized.