A Hybrid Approach Using Topic Modeling and Class-Association Rule Mining for Text Classification: the Case of Malware Detection
2018 IEEE 17th International Conference on Cognitive Informatics & Cognitive Computing (ICCI*CC)(2018)
摘要
We propose a novel general-purpose hybrid method comprising topic modeling and Class Association Rule Mining (CARM) for text classification in tandem. While topic modeling performs dimension reduction, association rule mining aspect is taken care by Apriori and Frequent Pattern(FP)- growth algorithms, separately. In order to illustrate the effectiveness of the proposed method, malware prediction using two publicly available datasets of API calls has been performed. The proposed model has generated highly accurate class association rules and Area Under the Curve (AUC) compare to the extant models in the literature. With the help of statistical significance test, it is concluded that the performances of both proposed hybrid models, i.e., topic modelina with FP-2rowth and Apriori, are same.
更多查看译文
关键词
Class Association Rule Mining,Text Classification,Topic Modeling,Latent Dirichlet Allocation,Malware Detection
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络