Research On Material Demand Forecast Of Manufacturing Enterprises Based On Deep Learning

With the increasing amount,types and scale of data,the requirements of data processing methods for enterprises are higher and higher.After it was formally proposed in 2006,deep learning has been favored by researchers from all walks of life.With a large number of data training,deep learning is better than traditional machine learning.In this paper,deep learning is applied to the material demand prediction of manufacturing enterprises.The time series data of sales demand and previous material demand are used as input,and the time series model and BP neural network are combined to train the data,and then the material demand value in the future period is output.At present,most enterprises base their material demand forecast on the ERP system.The main process is to generate the master production plan through the sales demand forecast,then generate the material demand plan through the decomposition algorithm,and finally determine the specific material purchase plan based on the bill of materials,inventory information and production capacity.Although the prediction method based on decomposition algorithm can analyze the internal structure of data and explain the source of material prediction clearly,it needs a lot of internal accounting to obtain data,which wastes a lot of time,and the prediction lead time is short,and the response time to manufacturing enterprises and their suppliers is too short,which ultimately results in the material supply can not meet the manufacturing requirements The production plan of the enterprise forces the enterprise to change the situation of production volume.Therefore,this paper takes the historical data of sales demand forecast and material information as input,establishes the model through BP neural network,taking into account not only the decisive factor of sales demand forecast,but also the material consumption in history,so as to achieve the purpose of making the forecast more accurate.In this paper,two models are used to predict the material consumption,which are CNN model and cnn-lstm model.By comparing the results of the two models with different independent variables,the prediction error is analyzed.The results show that when the independent variables include sales demand forecast,material consumption forecast and material attribute information,the prediction error of the two models is smaller and the result is better than that of the independent variables.At the same time,the prediction results of cnn-lstm model are better than that of CNN model.Using cnn-lstm model to predict material demand can solve two problems in the process of material demand prediction:(1)short lead time,(2)complex operation,data acquisition and accounting difficulties.The research results provide a more scientific solution for the material demand prediction of manufacturing enterprises.