Malicious Click Identification and Prediction Method Based on Optimized RF

To address the issue of malicious clicks affecting the accuracy of recommendation systems in e-commerce platforms, this study investigates the identification and classification algorithms for malicious click behavior. The research processes and selects features of the data through the SMOTE and the recursive feature elimination algorithm, and conducts model predictions based on the logistic regression, support vector machine, and random forest algorithms respectively. Through multiple evaluation indicators, the performance of the models is comprehensively assessed. The experiments show that among the three traditional models, the random forest model performs the best. On this basis, the genetic algorithm is used to optimize the parameters of the random forest model, and the results indicate that the optimized model has improved classification performance for the minority class. Finally, based on the actual scenarios of ecommerce platforms, relevant suggestions are provided from the aspects of optimizing recommendation systems and improving platform management strategies.