关键词: 混合式, 特征选择, 隐私保护, K-匿名, 搜索策略

Abstract: Abstract: In the era of big data, the protection of data privacy has become an issue that cannot be ignored. The K-anonymity algorithm is a classic method in the field of privacy protection. The data can reach the K-anonymity condition by generalizing and hiding the data. When hiding features, considering the privacy and classification performance of the data can be regarded as a special Feature selection method, namely K-anonymous feature selection. The K-anonymity feature selection method combines the characteristics of K-anonymity and feature selection to select a subset of K-anonymity features using multiple evaluation criteria. The filtering K-anonymous feature selection method is difficult to search for all candidate feature subsets that meet the K-anonymity condition, and cannot guarantee that the classification performance of the feature subset is optimal, and the encapsulated feature selection method has a large computational cost. Therefore, the article combines the characteristics of filtered feature ranking and encapsulated feature selection, improves the forward search strategy in the existing methods, and designs a hybrid (Hybrid) K-anonymous feature selection method, using classification performance as the evaluation criterion to select the subset of K-anonymous features with the best classification performance. Experiments on multiple public data sets show that the algorithm proposed in this paper can surpass the existing methods in classification performance, and the information loss is smaller.

Key words: hybrid, feature selection, privacy protection, K-anonymity, search strategy