Parameter Optimization of Local-Concentration Model for Spam Detection by Using Fireworks Algorithm

Abstract

This paper proposes a new framework that optimizes anti-spam model with heuristic swarm intelligence optimization algorithms, and this framework could integrate various classifiers and feature extraction methods. In this framework, a swarm intelligence algorithm is utilized to optimize a parameter vector, which is composed of parameters of a feature extraction method and parameters of a classifier, considering the spam detection problem as an optimization process which aims to achieve the lowest error rate. Also, 2 experimental strategies were designed to objectively reflect the performance of the framework. Then, experiments were conducted, using the Fireworks Algorithm (FWA) as the swarm intelligence algorithm, the Local Concentration (LC) approach as the feature extraction method, and SVM as the classifier. Experimental results demonstrate that the framework improves the performance on the corpora PU1, PU2, PU3 and PUA, while the computational efficiency is applicable in real world.