摘要
为提高销量预测的准确率,提出一种基于改进麻雀搜索算法优化LSTM的销量预测模型。首先在算法迭代过程中通过重心反向学习对个体进行变异,以增强算法跳出局部最优的能力;其次利用改进的算法对LSTM神经网络的超参数进行优化,解决了依靠主观经验选取超参数时存在精度不佳的问题;最后在原始销售数据的基础上加入大量零售数据等多个特征变量进行辅助预测,提高模型预测准确性。实验结果表明,该模型相较于其他模型具有更好的预测效果。
In order to improve the accuracy of sales forecasting,a sales forecast model based on LSTM optimized by improved SSA is proposed.Firstly,the centroid opposition-based learning is used to mutate individuals,so as to enhance the ability of the algorithm to jump out of the local optimal.Secondly,the improved algorithm is applied to the optimization of LSTM neural network hyperparameters,which solves the problem of poor accuracy when selecting hyperparameters based on subjective experience.Finally,a large number of characteristic variables,such as retail data,are added to the original sales data for auxiliary prediction to improve the prediction accuracy of the model.The experimental results show that this model has better prediction effect than other models.
作者
楼泽霖
郑军红
何利力
Lou Zelin;Zheng Junhong;He Lili(School of Computer Science and Technology,Zhejiang Sci-Tech University,Hangzhou,Zhejiang 310018,China)
出处
《计算机时代》
2023年第10期50-53,58,共5页
Computer Era
基金
浙江省重点研发“尖兵”攻关计划项目(2023C01119)。
关键词
LSTM
超参数优化
麻雀搜索算法
重心反向学习
销量预测
LSTM
hyperparameter optimization
sparrow search algorithm(SSA)
centroid opposition-based learning
sales forecasting
作者简介
楼泽霖(1998-),男,浙江诸暨人,硕士研究生,主要研究方向:数据智能;通讯作者:郑军红(1978-),男,浙江磐安人,博士,讲师,主要研究方向:商务智能、人工智能。
