基于SDAE-WDL的推荐策略优化研究
首发时间:2020-12-17
摘要:本文针对传统的推荐模型方法在处理海量高维度数据时效果不佳的问题,提出了一种融合栈式去噪自编码器(SDAE)和WDL(Wide& Deep Learning for Recommender Systems)的推荐系统。首先,利用栈式去噪自编码器对数据进行特征降维,实现从高维度到低维度的非线性转换;然后利用WDL模型对数据进行分类推荐,利用Movielen数据集的实验结果表明,与其他推荐模型方法相比,SDAE-WDL模型性能要优于其他方法,取得了更好的推荐效果。
For information in English, please click here
Research on Optimization of Recommendation Strategy Based on SDAE-WDL
Abstract:Aiming at the problem that the traditional recommendation model method is not effective when processing massive high-dimensional data, this paper proposes a recommendation system that combines stack denoising autoencoder (SDAE) and WDL (Wide & Deep Learning for Recommender Systems). First, use the stacked denoising autoencoder to reduce the feature dimension of the data to realize the non-linear conversion from high dimension to low dimension; then use the WDL model to classify and recommend the data, and the experimental results of the Movielen dataset show that it is different from other Compared with the recommended model method, the performance of SDAE-WDL model is better than other methods, and a better recommendation effect is achieved.
Keywords: Neural network recommendation system model fusion
基金:
引用
No.****
同行评议
勘误表
基于SDAE-WDL的推荐策略优化研究
评论
全部评论