Research And Application Of Excavator Sales Forecast Algorithm Based On Time Series

Excavator is an important production tool in construction infrastructure and other activities.Sales is a means of monetizing excavator equipment manufacturers.The analysis and prediction of excavator sales volume can effectively help decision-makers to perceive market trends in advance,and it is particularly important for realizing scientific inventory management and effective allocation of resources.Although data mining technology has been applied to sales forecast in e-commerce,retail industry,passenger car and other fields,it still uses expert prediction and other methods in the field of construction machinery such as excavator industry.Therefore,on the basis of the analysis of the factors affecting excavator sales,this paper collects necessary data,constructs effective features,studies the single-step excavator sales prediction algorithm and multi-step excavator sales prediction algorithm,and finally realizes the development of excavator sales prediction system.The main work contents are as follows:First of all,this paper fully investigate the business background of excavator sales forecast,analyzes the research status of sales forecast and time series at home and abroad.The main factors affecting excavator sales are real estate infrastructure,upgrading,macroeconomic,environmental protection factors and so on.The problem of excavator sales forecast is transformed into the problem of time series prediction and analysis in data mining,and the corresponding mathematical theoretical model is established.Next,the excavator sales forecast data set is established by cooperating with excavator manufacturers and data organizations.A periodic factor method is proposed to fill in the missing values of sales data,which takes into account both long-term trend and short-term fluctuation,and a logarithmic method is used to alleviate the long-tail distribution effect of sales volume to accelerate the convergence of the model.Through exporatory data analysis(EDA),the shallow data pattern in the data set is found,and a feature extraction framework for excavator sales forecast is designed.Aiming at the sales forecast of excavator in a single year and a single month,combined with the business background of excavator sales forecast,the prediction algorithm of three time series is studied and improved.It mainly includes the STL-ARIMA algorithm which decompose the sales series STL and then adopt ARIMA modeling to alleviate the instability problem after difference,the LGB-FILTER-BYS algorithm which integrates the feature selection method based on iterative idea and Bayesian automatic parameter tuning method on the basis of GBT algorithm,and hybrid input sales forecast algorithm based on CNN-LSTM.On this basis,the application of Stacking and average method for sub-model fusion,effectively improve the single model sales forecast effect.Aiming at the problem of excavator multi-month sales forecast,a multi-step time series prediction algorithm is studied.Using recursive strategy and direct strategy,the multi-step prediction problem is transformed into a single-step prediction problem,and the recursive multi-step prediction model and direct multi-step prediction model based on LGB-FILTERBYS are constructed.In order to enhance the correlation between prediction labels and alleviate the accumulation of multi-step prediction errors,a sequence generation sequence multi-step prediction model based on LSTM was constructed.Aiming at the problem that LSTM cannot accelerate parallel learning,a CNN-based sequential generation sequence multistep prediction model was built.Under the similar training time,CNN uses deeper network structure and achieves more accurate prediction results than LSTM.Finally,the excavator sales forecast system is designed and implemented.The system integrates data visualization results,time series feature calculation,single-step excavator sales forecast and multi-step sales forecast algorithm.This system provides clear information guidance for all kinds of decision-makers in excavator manufacturers.