Research On Sales Of Dishes Forecast Based On Deep Learning And Combination Model

Due to the different formats of restaurants in the catering industry,the forecasting research of dish sales is also relatively complex.In this paper,we select the representative formats: barbecue and Chinese food,and establish dishes sales forecasting models suitable for these two formats respectively.In the forecasting model of barbecue industry,a new forecasting model IPSO-LSTM-GAN is proposed from the perspective of influencing factors.This part selects the operating data of a barbecue restaurants for research.Firstly,an improved particle swarm algorithm is proposed and four different optimization models are established,namely PSO-BP,IPSO-BP,PSO-LSTM and IPSO-LSTM.The empirical results shows that the predictive index effect of IPSO-LSTM model is the best.Secondly,Back Propagation(BP)network is used as the discriminator and Long Short Term Mermory network(LSTM)as the generator to improve the original Generative Adversarial Networks(GAN)network structure and establish the GAN model.The results show that the GAN model can accurately predict the sales of dishes.Finally,a combined forecasting model based on IPSO-LSTM and GAN is established by using non equal weight average method,which is called IPSO-LSTM-GAN.The final empirical results show that: Compared with the IPSO-LSTM and GAN models,the predictive index root-mean-square error(RMSE)of the proposed IPSO-LSTM-GAN is lower,which can more accurately predict the change of food sales.In the forecasting model of Chinese food format,two combined forecasting models are proposed from the perspective of time series data.This part selects the time series data of a Chinese food restaurants for empirical research.Based on the data preprocessing,the first type of combination model establishes a total of 5 combination models,and the second type of combination model establishes a total of 12 combination models.The final comparative experimental results show that the proposed the second type of combination model based on the combination of equal weight method and error reciprocal method has a higher prediction accuracy.At the same time,in order to improve the supply efficiency of food sales in the catering industry,a scheduling model based on constraint programming was established in the end of this paper,and using Google's ortools tool to obtain the optimal scheduling solution,and then providing a comprehensive reference for catering enterprises.