基于最小二乘法的混合推荐模型研究

Research on hybrid recommendation model based on least square method

针对单一推荐模型在电影推荐过程中无法同时利用推荐系统中的隐式信息和显式信息所导致的推荐不准确以及冷启动等问题,提出了一种基于最小二乘法的混合推荐模型。该模型首先通过基于内容的推荐算法和协同过滤推荐算法分别进行单一模型的推荐,然后对单一推荐模型所产生的推荐结果动态地调整权重进行数据拟合,再将所产生的拟合数据进行最小二乘运算,减小整体预测误差,从而得到最终的推荐结果。最后使用MovieLens 100k和MovieLens 1M这两种公开的电影数据集对该模型进行验证并与其他几种模型进行比较。实验结果表明,所提出的基于最小二乘法的混合推荐模型在精确率、召回率和F值等评价指标上都优于目前几种传统的推荐模型,所造成的预测误差相较于目前几种传统推荐模型也更小。

In order to solve the problem that a single recommendation model cannot simultaneously utilize the implicit and explicit information in the recommendation system in the process of movie recommendation,which may cause inaccurate recommendation and cold start,a hybrid recommendation model based on the least square method is proposed.In the model,the content-based recommendation algorithm and the collaborative filtering recommendation algorithm are adopted respectively to conduct a single model recommendation,the weights of the recommendation results generated by the single recommendation model are dynamically adjusted to realize data fitting,and then the least square operation for the fitting data is performed to reduce the overall prediction error,so as to get final recommendation result.In the end of the article,two common movie datasets(MovieLens 100k and MovieLens 1M)are used to verify the model and compare it with the other models.The experimental results show that the proposed hybrid recommendation model based on the least square method is superior to the currently-used traditional recommendation models in terms of precision,recall and F-measure,and its prediction error is also smaller than those of the currently-used traditional recommendation models.