基于多元正态概率模型的贝叶斯概率矩阵分解研究

发布时间：2018-05-31 13:42

本文选题：电子商务 + 大数据　；参考：《科技通报》2017年09期

【摘要】：传统的概率矩阵分解技术,忽视了信息消费者间的信任关系和关注关系,使其推荐性能和推荐质量不断下降,针对以上问题,提出基于多元正态概率模型的贝叶斯概率矩阵分解算法,以多元正态概率模型作为先验分布,实验中通过计算Gibbs sampling过程中迭代次数来达到数据的稀疏性。在联合概率未知和条件概率易得等情况下,引用Gibbs sampling技术进行计算,实验中引用MAE、RMSE两种方法进行误差评价。实验结果表明:在稀疏矩阵的检测中,改进的贝叶斯概率矩阵分解算法的预测精密度更加稳定,对缓解矩阵稀疏性问题更加有效。
[Abstract]:The traditional probability matrix decomposition technology neglects the relationship of trust and concern among information consumers, which causes the performance and quality of recommendation to decline. A Bayesian probability matrix decomposition algorithm based on multivariate normal probability model is proposed. The multivariate normal probability model is used as the prior distribution and the data sparsity is achieved by calculating the iterations in the Gibbs sampling process. Under the condition that the joint probability is unknown and the conditional probability is easy to obtain, the Gibbs sampling technique is used to calculate the error, and the mae RMSE method is used to evaluate the error in the experiment. The experimental results show that the prediction precision of the improved Bayesian probability matrix decomposition algorithm is more stable in the detection of sparse matrix, and it is more effective to alleviate the sparse problem of matrix.
【作者单位】：湖南大学新闻传播与摄影艺术学院;开封大学实验实训中心;
【分类号】：O212.8

【相似文献】