Petroleum Lithology Discrimination Based on PSO-LSSVM Classification Model

This paper proposes an algorithm which combines Particle Swarm Optimization (PSO) with Least Squares Support Vector Machines (LSSVM) to identify lithology by using well logging data. First of all, PSO is used for optimizing the main parameters of LSSVM, and then by using the optimized parameters to obtain a better PSO-LSSVM classification model which can be used to identify lithology with logging data. Compared with the traditional SVM model based on cross-validation and a single hidden layer of BP neural network model, the new PSO-LSSVM method can accurately describe the nonlinear mapping relationship between the well logging data and the lithology categories. The experimental results show that a higher precise identification can be got and the automation of the algorithm can also be improved.