Abstract:

Feature selection is to select feature subsets from the original feature set, and it can reduce the dimension of feature and redundant information, so as to improve the accuracy of classification. In order to achieve this effect, a new feature selection algorithm is proposed in this paper. The algorithm uses the enhanced fireworks algorithm after discretization to search the feature subset. At the same time, the feature subset and the constraint conditions after penalty factor processing are integrated into the objective function. Then the data of the feature subset are trained and predicted by the [kNN] classifier. Finally, the accuracy of classification is tested by 10-fold cross validation. Compared with the guided fireworks algorithm, fireworks algorithm, bat algorithm, crow search algorithm and adaptive particle swarm optimization algorithm, the simulation results using UCI data show that the overall performance of the proposed algorithm is better than that of the other five algorithms.

Key words: discrete enhanced fireworks algorithm, feature selection, dimension reduction, classification, [k]-Nearest Neighbor[(kNN)]

摘要：

特征选择是从原始特征集中选取特征子集，并且降低特征维度和减少冗余信息，从而达到提高分类准确度的效果。为了达到此效果，提出了新的特征选择算法。该算法使用经过离散化处理之后的增强烟花算法来搜索特征子集，同时将特征子集和经过惩罚因子处理之后约束条件融入到目标函数中，然后将搜索到的特征子集的数据放到[kNN]分类器进行训练和预测，最后使用十折交叉验证来检验分类的准确性。使用UCI数据进行仿真实验，仿真结果表明：与引导型烟花算法、烟花算法、蝙蝠算法、乌鸦算法、自适应粒子群算法相比，所提算法的总体性能优于其他五种算法。

关键词: 离散型增强烟花算法, 特征选择, 降维, 分类, [k]近邻[(kNN)]