A non-iterative approach to kernel logistic regression for imbalanced data

A non-iterative kernel logistic regression learning method for severely imbalanced data was proposed. The method is an improvement on the iterative weighted least squares method for classical kernel logistic regression. It not only reduces the computational burden caused by iteration, but also utilizes the knowledge of the ratio of the benchmark category, and can avoid problems normally encountered when processing imbalanced data such as undersampling, oversampling and cost-sensitive learning. Thus, this method enables the efficient and fast modelling of kernel based logistic regression in the case of large-scale imbalanced data, through the construction of a robust modified least square logistic classifier. Theoretical research indicates that the proposed method has some excellent properties, and simulation research and empirical studies show that its classification effect is good.