Research on Dynamic Cost-Sensitive SVM Classifier based on Chaos Particle Swarm Optimization Algorithm

Abstract

In order to improve the performance of Support Vector Machine (SVM) classifier for imbalanced data, this paper proposes dynamic cost-sensitive SVM classifier based on chaos particle swarm optimization (CPDC_SVM). IntroduceSupport vector machine (SVM) based on statistical learning theory is a new machine learning method, and it can well resolve such practical problems as nonlinearity, high dimension and local minima. As the mainstream statistical machine learning methods, Support Vector Machine has good performance, and is applied in the image detection, biometrics, and information security. There are classic one-against-one SVM, one-against-rest SVM, the directed acyclic graph SVM, and the decision tree SVM. However, the existing classical SVM classifications are sensitive to noise and aren't suitable to imbalance dates, and it focus overmuch the performance of the majority class in data. How to design a new SVM model for fitting the imbalance data is a hot spot research. There are some improved SVM models for the unbalanced data classification. Gonzalez introduced a new support vector machine GSVM, which is specially designed for bi-classification problems where balanced accuracy between classes is the objective [1]. Mordelet proposed a new method for PU learning with a conceptually simple implementation based on bootstrap aggregating (bagging) techniques [2]. Zhao proposed a weighted maximum margin criterion to optimize the data-dependent kernel, which makes the minority class more clustered in the induced feature space [3]. The paper [4] proposed the structural imbalance SVM, [5] and [6] improved directly C-SVM model, and used the different penalty factors for the different classes. Liu adopt AUC to measure the performance of classifier. In order to overcome the noise to influence the classifier

Acceleration harmonic estimation for a hydraulic shaking table by using particle swarm optimization

“…Sedighizadeh and Kashani (2015) made use of Tribe-PSO algorithm, multi-layered and multi-phased of hybrid PSO model to identify parameters of proton exchange membrane fuel cell model. Zhang (2014) investigated a dynamic cost-sensitive support vector machine classifier based on chaos PSO to improve its performance for imbalanced data. Alsmadi et al (2014) utilized PSO to develop a model-order reduction technique with the advantage of critical frequency preservation capability.…”