基于宽度学习和协同过滤的在线信息推荐算法

ONLINE INFORMATION RECOMMENDATION SYSTEM BASED ON BROAD LEARNING AND COLLABORATIVE FILTERING

基于深度学习算法的协同过滤推荐方法由于准确率高而被广泛应用,然而采用深度学习算法的模型训练过程复杂耗时,泛化能力弱,使得在线推荐系统的效率大幅降低。为了在保证在线推荐精度的前提下提升推荐系统的效率,提出建立基于协同过滤的信息推荐系统,应用宽度学习模型预测用户偏好,实现在线计算。通过宽度学习在公共电影数据集MovieLens的数据实验,并与深度神经网络模型和轻量化梯度提升模型的实验结果在精度和效率两个层面进行对比,结果表明基于宽度学习模型的推荐系统可以有效提高模型训练的速度,并实现较高的准确率。

The collaborative filtering recommendation method based on deep learning algorithms has been widely concerned due to its high accuracy.However,the training process of the model applying these deep learning algorithms is complicated and time-consuming,and the model generalization ability is weak.These issues decrease the efficiency of the recommendation system.To improve the efficiency of the recommendation system on the premise of ensuring the accuracy of online recommendation,this paper establishes an information recommendation system based on the collaborative filtering methods.It predicted user preference using the broad learning model to achieve online calculation.Through the experiment of a public movie data set MovieLens based on the broad learning model and comparisons with deep neural network and light gradient boosting machine on accuracy and efficiency,the experiment result shows that the proposed information recommendation system can effectively improve the training speed while achieving relatively high accuracy.