Equipment Spare Parts Demand Forecasting For Construction Companies Based On Prophet And LSTM Models

As one of the pillar industries of China’s modern economic development,the construction industry is also booming with the development of China’s economy and society.Since 2013,the level of capacity of China’s construction industry has significantly increased,the scale of each enterprise has steadily grown,the amount of work undertaken has steadily increased,and the technical strength of the entire construction industry has significantly improved.However,due to the characteristics of the construction industry,such as many construction points,wide coverage and long construction cycle,its labor costs continue to rise and material costs remain high.In order to reduce construction costs,inventory control has now become the main means to enhance the efficiency of enterprise material management and improve revenue,and it is of practical significance to study the demand forecasting and control of material and materials in the construction industry.This paper will use the relevant theories of demand forecasting and the relevant methods in the field of time series forecasting to forecast the demand for spare parts of machinery and equipment in construction enterprises and provide new ideas and solutions for the demand control of equipment spare parts in construction enterprises.This paper analyzes and compares the characteristics of various forecasting models on the basis of summarizing the theories and methods of inventory management and demand forecasting at home and abroad.Among them,Prophet model,which has the characteristics of fast fitting,flexible adaptability and strong interpretation,is applicable to the data of business behavior with obvious intrinsic rules,data with continuous historical trends,LSTM(Long Short Term Memory)neural network model has good prediction on the unstable time series with more fixed components It is suitable for non-linear time series.In the construction enterprise equipment spare parts demand forecasting,the historical data of the research target contains not only linear relationships but also nonlinear components.In order to improve the practicality and prediction accuracy of the model,the Prophet-LSTM combined prediction model is constructed based on the historical consumption characteristics and consumption of equipment spare parts,and the material demand management approach of construction enterprises.non-linear errors in the prediction results.The Tensorflow framework and Facebook Prophet Python plugin are used to build the Prophet-LSTM combined model runtime environment under Anaconda integrated environment.The 13-month consumption data of equipment spare parts of a project department of enterprise A were selected as the original data,and the data set was divided into a training set and a test set,where the training set was the data of a whole year period and the test set was the data of a month period;then the data in the test set were predicted using Prophet and Prophet-LSTM combined model,and the mean absolute error(MAE)and root mean square error(RMSE)were calculated.RMSE,and a comparative analysis of the prediction errors was performed.The results show that the combined Prophet-LSTM model effectively reduces the error of equipment spare parts demand forecasting by more than 10% compared with the single model of Prophet,and improves the forecasting accuracy,thus providing a reference basis for the enterprise machinery and equipment spare parts procurement scheme.