An Improved Interval-Valued Campbell-Bennett OCC Model Based on PSO

This paper proposes a method for parameter optimization of the interval-valued Campbell-Bennett model proposed by Utkin and Chekh using the Particle Swarm optimization (PSO) method. Campbell-Bennett model is a OneClass Classification (OCC) model, whose dual problems can be expressed as a set of simple linear programming problems. Utkin uses the triangular kernel function approximation to replace the Gaussian kernel function in the Campbell-Bennett model, and thus obtains a finite set of simple linear optimization problems for processing interval-valued data. However, the complexity of this method is too high, making it difficult to optimize model parameters. In order to avoid this high complexity, we use a special method to optimize the model parameters with PSO and verify it with numerical experiments.