一种基于受限波尔兹曼机的推荐算法

A Recommendation Algorithm Based on Restricted Boltzmann Machine

在数据量过于庞大的情况下,RBM模型所输出的推荐结果会比较宽泛。此外,目前众多的协同过滤算法无法对巨大的数据集进行更好的处理。所以,尝试通过深度学习来对个性化推荐进行加强,指出把受限波尔兹曼机与隐含因子模型相结合的混合推荐方法。首先用RBM算法生成候选集,并对候选集的稀疏矩阵进行评分预测,然后使用LFM对候选结果进行排序,进而选择最优方案进行推荐。使用大型公开数据集对本文算法进行反复验证,通过测试可以看出,相比较于传统的推荐模型,本文所提阐述的方式能够有效提高评分预测的精准度。

In the case where the amount of data is too large, the recommended results output by the RBM model will be broader. Besides, many collaborative filtering algorithms currently do not handle large data sets better. So, we try to use the deep learning technology to strengthen the personalized recommendation model. We propose a hybrid recommendation model combining the bound Boltzmann model and the hidden factor model. First, we use the RBM algorithm to generate candidate sets, and score the sparse matrix of the candidate set. Then we use the LFM model to sort the candidate results and select the optimal solution for recommendation. The hybrid model is validated using used large public datasets. It can be seen from the verification that compared with the traditional recommendation model, the proposed method can improve the accuracy of the score prediction.