Abstract: In order to solve the problems of insufficient accuracy of the existing landslide susceptibility evaluation model and the limitations of a single decision-making model, an integrated PSO-GA-RF integrated model with an integrated intelligent combination algorithm to optimize RF is proposed. The well-known filtering feature selection method(Relief-F algorithm)is used to rank the weights of landslide-causing factors, eliminate redundant features, and optimize the classification results, thus reducing the problem of relying on subjective judgments to extract the influencing factors, and lowering the human error. The PSO-GA-RF integrated model combines the advantages of multiple algorithms to optimize the parameters of the RF model, which simplifies the tedious process of parameter selection and reduces the error. The experimental results show that the PSO-GA-RF integrated model outperforms the RF and GA-RF models in terms of prediction performance and efficiency.

Key words: landslide, vulnerability, influence factor, Relief-F, PSO-GA-RF