基于多标签机器学习的环境适老化改造应用研究

Study on the application of environment-oriented aging transformation based on multi-label machine learning

针对环境适老化改造推荐方案中改造项具有先后顺序的情形,将校准标签排序与多标签k近邻算法相结合,提出适于适老化改造推荐的多标签学习算法。首先引入校准标签约束不相关项,文章将多标签排序问题进行转化为标签成对比较的多分类问题,然后利用ML-k NN进行多标签分类,最后重构标签对关系矩阵,取统计票数大于校准标签的标签项为相关标签并根据票数进行排序。通过实际老人数据集进行测试,与传统的成对比较排序法与校准标签法进行比较,结果表明校准标签排序与多标签k近邻算法相结合汉明损失更小,平均精度更高,且保持较低的错误率,更适合于适老化改造方案推荐。

Aiming at the ranking of reconstruction items in the recommendation scheme of environment-oriented aging transformation,the calibration label ranking is combined with the Multi-Label k-Nearest Neighbor algorithm,and a multi-label learning algorithm suitable for adapting to aging transformation is proposed.First,the calibration label is introduced to constrain unrelated labels,and the multi-label sorting problem is converted into a multi-classification problem in which pairs of labels are compared.Then ML-kNN is used for multilabel classification.Finally,the label pair relationship matrix is reconstructed and the votes of relation labels should be greater than the calibration and sorted according to the number of votes.The actual elderly dataset was tested and compared with the traditional Ranking by Pairwise Comparison method and Calibration Label Ranking method.The results show that the combining Calibration Label Ranking with the Multi-Label k-Nearest Neighbor algorithm cause Hamming loss smaller,the average precision higher,and the 1-error relatively low,which is more suitable for the environment-oriented aging transformation.