Sales Forecast Of University Canteen Window Based On The Two-layers Model Of Sales Forecast

With the advent of the era of big data,people begin to realize that the data accumulation of catering industry has not been fully utilized.As far as colleges and universities are concerned,education big data and education data mining are getting more and more attention.As an important part of education big data,the campus card swiping data of each canteen window has been widely analyzed,mined and utilized.Now food waste has become one of the common problems in the world.In order to avoid the severe food waste problem,and respond the call for chairman Xi’s" clear your plate " campaign,it is very important to establish the accurate sales forecast model based on campus ID card data from each window in college dining rooms.The existing sales forecast mainly comes from the experience of the staff,whose accuracy is poor.Meanwhile it is more for long-term forecast.The purpose of this thesis is to use the dining room window card history data to set up an accurate short-term prediction model,for canteens window of preparing raw materials,marketing activities to provide big data service and support.As the main dining places of college students,college canteens are more perfect than off-campus catering industry to some extent,but there are still some problems because most of the management systems are made by relevant personnel according to experience.For example,the problem of food waste and the problem of insufficient food supply,the former not only causes waste but also reduces the profit of the canteen window,while the latter may affect the dining experience of students,leading to the decrease of student satisfaction in the dining rooms.If you can make full use of the student card consumption data in the college canteen window,use statistical methods and data analysis and mining techniques to establish an accurate window sales forecast model,you can effectively avoid the above two major problems and fully improve the student satisfaction at the canteen window.In this thesis,the author collected ID card daily total sales data from different canteen windows in Central China Normal University and regard them as time series data,First of all,different sources of canteen window data have different characteristics from descriptive and visualization analysis.In addition,as this thesis mainly studies short-term prediction,different time segments of time series data will also lead to different characteristics of sample data.In the second step,the sliding segmentation is carried out for the original data to obtain the sample timing data.The third step,the author adopted on all sample data based on Seasonal AutoRegressive Integrated Moving Average model and the Xgboost regression model based on machine learning model prediction,indicating that different algorithms model under the different characteristics of the sample data is different.Generally speaking,the effect of the different algorithms model is suitable for prediction of different characteristics of the sample data.The basic prediction model still has a lot of room for improvement,and it does not have good scalability and iterative improvement.Therefore,this thesis puts forward a two-layers model of canteen window sales prediction based on the classification model after extracting statistical features from sample time series data,which can effectively improve the prediction accuracy.The two-layers sales forecast model established based on this experiment has a high prediction accuracy rate,which has better solved the problem of the logistics center in the canteen window sales forecasting,and effectively provided guidance for the business activities of the canteen.In addition,the model has the ability to automatically learn time series data features and automatically select basic prediction models,which can better face large-scale,multi-feature time series prediction problems,and can be well extended to time series prediction tasks in other scenarios.Such as financial sequence forecasting,tourism flow forecasting,etc.