基于全球恐怖主义数据库的特征选择方法研究

恐怖主义被称为现代人类社会之癌,是世界各国政府和人民面临着的重大的挑战,应该引起全人类的重视。在使用全球恐怖主义数据库中的数据对恐怖主义活动进行研究时,从高维数据中提取关键的特征,是反恐研究中的重点和难点。针对全球恐怖主义数据库中特征的高维性、冗余性和数据不完整性的特点,分别采用最小冗余最大相关算法(mRMR)、基于支持向量机的递归删除算法(SVM-RFE)和基于随机森林的特征选择算法进行特征筛选与提取。利用K-近邻(KNN)分类器其对上述特征选择方法进行降维结果分析和分类结果比较。实验结果表明,特征选择算法不仅能提高分类性能还能提高分类效率,并且基于支持向量机的递归删除算法(SVM-RFE)选择的特征子集在预测恐怖主义活动时准确率更高。

FEATURE SELECTION METHOD BASED ON GLOBAL TERRORISM DATABASE

Terrorism, known as the cancer of modern human society, is a major challenge faced by governments and people around the world and should be brought to the attention of all mankind. When using the data in the global terrorism database to study terrorist activities, extracting key features from high-dimensional data is the focus and difficulty in counter-terrorism research. According to the characteristics of high dimensionality, redundancy and data incompleteness in the global terrorism database, we adopted minimum-redundancy maximum-relevancy(mRMR), recursive feature elimination based on support vector machine(SVM-RFE) and the feature selection algorithm based on random forest respectively to screen and extract features. K-nearest neighbor(KNN) classifier was used to analyze the dimension reduction results and compare the classification results of the above feature selection methods. The experimental results show that the feature selection algorithm can improve not only the classification performance but also the classification efficiency. And the feature subset selected by SVM-RFE has higher precision when predicting terrorist activities.