基于销量预测的LSTM模型优化
Optimization of LSTM Model Based on Sales Volume Prediction
唐静芸 1郗鑫 1赵鹏1
作者信息
- 1. 太原师范学院计算机科学与技术学院,山西晋中 030619
- 折叠
摘要
为了解决销量预测精度不准确以及数据集贫乏的问题,实现结果更精准、效率更高的销量预测,提出一种基于特征工程的长短时记忆(LSTM)神经网络销量预测模型.LSTM模型是经过循环神经网络(RNN)模型优化后的一种特殊结构,适用于处理与时间序列相关的问题.通过特征工程提取、筛选、衍生出新的特征序列与原始特征序列融合,并结合LSTM算法进行销量预测,提高了 LSTM模型预测的精准度.以Rossmann店铺的实际销售数据为例,证实了经过优化后的特征序列集融合LSTM模型进行预测,在提高预测效率的同时也提升了预测结果的精准度.
Abstract
In order to solve the problems of inaccurate sales prediction accuracy and poor data-set,and achieve more accurate and efficient sales prediction results,a feature engineering based Long Short Term Memory(LSTM)neural network sales prediction model is proposed.The LSTM model is a special structure optimized by a recurrent neural network(RNN)model,suitable for dealing with time series related problems.By extracting,filtering,and deriving new feature sequences through fea-ture engineering and fusing them with the original feature sequences,and combining them with LSTM algorithm for sales prediction,the accuracy of LSTM model prediction has been improved.Taking the actual sales data of Rossmann store as an example,it has been confirmed that the optimized feature se-quence set fused with LSTM model for prediction not only improves prediction efficiency but also en-hances the accuracy of prediction results.
关键词
LSTM模型/销量预测/特征工程Key words
LSTM model/sales forecast/feature engineering引用本文复制引用
出版年
2024
NETL
NSTL
万方数据