融合特征提取的LMM模型在电商企业营销数量预测中的应用研究

Research on the Application of LMM Model with Fusion Feature Extraction in Forecasting the Marketing Quantity of E-commerce Enterprises

准确的销售预测可以帮助电商企业制定合适的营销计划,以减少业务损失。但传统的预测模型对如今大数据发展迅速的时代来说,已经不能满足电商企业的需求。基于特征提取,对销售数据进行傅里叶变换处理,构建一种线性混合预测模型(Linear Mixed Model),应用至电商企业营销数量的预测中。开展实验对预测模型进行效用分析,实验表明,神经网络、支持向量机模型的MAE为0.04927154,0.04847517,而基于特征提取的线性混合模型的MAE更小,为0.04032,说明模型质量和性能更好。涉及RMSE、MAE、RMSE和RMSPE评估指标时,三种模型误差相差均在0.1以内。电商营销数量数据具有时间序列特性,但是线性混合模型在解决时间序列问题方面更有效,因此能够更加有效地处理电商数据。

Accurate sales forecasts can help e-commerce enterprises develop appropriate marketing plans to reduce bus-iness losses.However,traditional prediction models are no longer sufficient to meet the needs of e-commerce enterpri-ses in the era of rapid development of big data.This study is based on feature extraction and Fourier transform processing of sales data to construct a Linear Mixed Model for predicting the marketing quantity of e-commerce enterprises.Con-duct experiments to analyze the utility of the prediction model.The experiments show that the MAE of the neural network and support vector machine models is 0.04927154 and 0.04847517,while the MAE of the linear hybrid model based on feature extraction is smaller,0.04032,indicating better model quality and performance.When it comes to the evaluation indicators of RMSE,MAE,RMSE,and RMSPE,the errors of the three models are all within O.1.E-commerce mar-keting quantity data has time series characteristics,but linear mixed models are more effective in solving time series problems,and therefore can process e-commerce data more effectively.