The Research On Sales Forecasting Model Based On Multidimensional Grey Model And Neural Network

Sales forecasting is significantly important for the enterprises.Accurate sales forecast can help the enterprises to develop right marketing strategies to reduce business losses and increase the profitability.However,owning to short product life cycle and a lot of influential factors,the historical sales data of fashion products are always few and volatile.So,it is hard to accurately predict.Acording to the characteristics of online sales,the influence factors of sales containing product itself and external environment were ananlyzed.The sales forecasting algorithm such as EELM,ARIMA,neural network,GM(1,1)and GM(1,N)had been compared.New forcast moldel IGM(1,N)based on GM(1,1)and GM(1,N)had been proposed.It can make the sales sequence satisfy the smoothing and exponential smoothing conditions.Besides,the relevant factors that influencing sales were also taken into consideration.The forecast accuracy is higher.Furthermore,hybrid intelligent algorithm based on the combination of IGM(1,N)and neural network has been proposed.During the prediction,the historical date was pre-treated.Grey relational grade analysis on relationship between influential factors and product sales was applied to determine major influences.The hybrid intelligent algorithm was used to forecast five data series.The performance of the proposed hybrid intelligent algorithm was evaluated by comparing with the actual value and assessing the errors.Results indicate that hybrid intelligent algorithm is feasible and has better forcasting performance than that of EELM,ARIMA,GM(1,1)and IGM(1,N)algorithm.The the mean absolute percent(MAPE)of sales sequences forecast is always around 24%.The major innovations are as follows:1)The IGM(1,N)algorithm has been proposed.The control factors can be added to the system characteristic sequence to make the sequence satisfy the requirements of the gray prediction method.Sales sequences influenced by multi factors can be predicted.2)The additional factors are added to make sure the sequences satisfy the requirements of the grey prediction method when IGM(1,N)was used to process the original sequences.3)Hybrid intelligent algorithm based on the combination of IGM(1,N)algorithm and neural network has been proposed.This algorithm can obtain high prediction accuracy and has better performance than EELM?ARIMA?GM(1,1)?IGM(1,N)algorithm.