混合蒙特卡罗搜索的特征选择算法的优化

特征选择是机器学习和数据挖掘中处理高维数据的初步步骤,通过消除冗余或不相关的特征来识别数据集中最重要和最相关的特征,从而提高分类精度和降低计算复杂度。文中提出混合蒙特卡罗树搜索特征选择算法(HMCTS),首先,根据蒙特卡罗树搜索方法迭代生成一个初始特征子集,利用ReliefF算法过滤选择前k个特征形成候选特征子集;然后,利用KNN分类器的分类精度评估候选特征,通过反向传播将模拟结果更新到迭代路径上所有选择的节点;最后,选择高精度的候选特征作为最佳特征子集。仿真结果表明,对比HPSO-LS和MOTiFS算法,HMCTS算法具有良好的可扩展性,且分类精度高。

Optimization of feature selection based on hybrid Monte Carlo Tree

Feature selection is a preliminary step in processing high-dimensional data in machine learning and data mining.It eliminates redundant or unrelated features to identify the most important and relevant features in the data set,thereby improving classification accuracy and reducing computational complexity.In this paper,a hybrid feature selection based on Monte Carlo tree search(HMCTS) is proposed.Firstly,an initial feature subset is generated by iteratively according to the Monte Carlo tree search method.The ReliefF algorithm is used to filter and select the previous feature to form a candidate feature subset.Then,the candidate features are evaluated by the classification accuracy of the KNN classifier,and the simulation results are updated to all selected nodes on the iterative path by backpropagation.Finally,the candidate features with high-precision are selected as the best feature subset.The simulation results show that the HMCTS algorithm has good scalability and high classification accuracy compared with HPSO-LS and MOTiFS algorithms.