多标签网页的粗糙集PNN高斯块植入期望排序推荐

GRPNN:Rough set PNN Gauss block based embedded expectation sorting algorithm for multi-label web recommendation

针对多标签网页推荐算法中存在信息不精确及新增信息较多,传统精确算法效果不理想的问题,提出一种多标签网页的粗糙集概率神经网络高斯块植入期望排序推荐方法。针对信息不确定性,利用粗糙集理论改进传统的概率神经网络模型,使之适合处理信息非确定性问题;针对固定概率神经网络在处理多标签网页推荐问题时,存在覆盖率差,结构冗余较大,对新增标签信息无法快速识别的问题,利用高斯块植入期望排序方式,构建概率功能块的合并、添加和删除操作,提高预测精度,降低计算复杂度,解决新增信息预测的实时性问题。在雅虎多标签数据集实例中的实验对比结果表明,该算法具有更高的计算精度和效率。

To solve the problem of imprecise,unsteady and incomplete information in multi label web page recommendation algorithms,and that the traditional accurate algorithms are not ideal,the rough set PNN Gauss block based embedded expectation sorting algorithm for multi label web recommendation was presented.For the problem of uncertainty of information,the rough set theory was used to improve the traditional probabilistic neural network,so that it was suitable for dealing with the problem of information uncertainty.To solve the problem of poor coverage and redundant structure in multi label page recommendation using fixed probabilistic neural network,the Gaussian block implantation expectation sorting algorithm was used to construct the merging,adding and deleting operations,improving the prediction accuracy and reducing the computational complexity.The experimental comparison of the multi label data sets in YAHOO shows the proposed algorithm has higher computational accuracy and efficiency.