基于扩展朴素贝叶斯与谱聚类的混合推荐算法的研究

A hybrid recommendation algorithm based on extended naive Bayes and spectral clustering

日常较为常用的推荐系统为协同过滤,其存在优势的同时也面临着很大的问题,因此需要寻求最优化的方式对其大量的数据进行分析。本文在常用的协同过滤算法之上,提出了一种基于扩展朴素贝叶斯与谱聚类的混合推荐算法,该混合推荐算法的原理是:首先将原本是数据的聚类问题通过谱聚类方法,转化为图的问题,通过此方法找到相近的数据,减小数据处理量;其次,结合扩展朴素贝叶斯算法,建立数据模型来预测数据,降低对时变数据处理的复杂性;最后根据用户对数据的兴趣变化,对模型进行局部调整,减少模型更新的复杂性。

The commonly used recommendation system is collaborative filtering,which has advantages but is also faced with great problems.Therefore,it is necessary to seek an optimized way to analyze a large amount of data.On collaborative filtering algorithm,this paper puts forward a simple based on extended bayesian spectral clustering and hybrid recommendation algorithm,the principle of the hybrid recommendation algorithm,originally is the data clustering problem through spectral clustering method,and transformed into the problem of figure,find the relevant data through this method,reduces the amount of data processing;Secondly,by combining with the extended naive Bayes algorithm,the data model is established to predict the data,which reduces the complexity of the time-varying data processing;Finally,according to the change of user's interest in data,the model is locally adjusted to reduce the complexity of model updating.