Research On Clothing Sales Forecast Based On Prophet-RF Combination Model

With the rapid development of Internet information technology and the rise of big data,cloud computing and other technologies,mobile payment and online transactions emerge as The Times require.The improvement of technology not only facilitates people's life,but also improves the production efficiency of enterprises and optimizes the management means of enterprises.Not only that,but also gave birth to a large number of sales data,which contained many important knowledge information.If an enterprise can make good use of these big data information,it can analyze the future sales trend with historical sales data and give forecast value through excellent algorithm model,which is undoubtedly the core means of business competition for enterprises.Through product marketing momentum can reasonable distribution of the production plan,make correct marketing strategy,can reduce caused by production and does not match the actual sales of out of stock or inventory,to avoid for the economic consequences of the not enough grasp of market,so the importance of the sales forecast for enterprise business decision is self-evident.In view of the explosion of sales data in the clothing industry,marketing market grasp is not accurate enough,and the imbalance of supply and demand of offline stores occurs,this paper proposes to use machine learning random forest model to improve the time series model,and the accuracy rate is not accurate due to the imperfect extraction of nonlinear information,and genetic algorithm is used to optimize the parameters of random forest model to make the combined model prophet RF prediction accurate Degree is more accurate.Based on ARIMA model,prophet model and prophet prediction model based on Stochastic Forest algorithm optimization are established.The main work to build a high-precision prediction model is as follows:(1)Analyzed the characteristics of the original data,screened the relevant influence variables,preprocessed the original data,scaled compressed the sequential data,reduced the fluctuation of the data,and carried out multi-dimensional merging of the relatively pure sales data according to the business requirements.(2)Through the analysis of the characteristics of historical sales data,the random forest model was selected,and the genetic algorithm was used to optimize the important parameters and training model of the random forest model.(3)Information on the nonlinear time series model can't better refining,and random forest model for nonlinear information has the property of strong ability of learning and using the optimized random forests model correction Prophet model to predict residual,absorbing time series model and the model of machine learning their respective advantages,build to get the optimum combination model based on Prophet-RF.(4)Using scientific indicators to demonstrate the performance of the equilibrium model by experiments.The experimental results are as follows: under the same experimental conditions,the average absolute error percentage of the traditional time series ARIMA model is 10.83%.The average absolute error percentage of Prophet model is 7.18%.The average absolute error percentage of the optimal combination model based on Prophet-RF is4.15%.The experimental data show that the traditional time series model cannot meet the accuracy requirements well in this business environment,while the combination model based on Prophet-RF optimization can not only reduce the variation risk probability of single model that is vulnerable to external factors,but also improve the prediction accuracy to some extent.