Sparse Optimization for Poisson Regression Based on GPGN Algorithm

Poisson regression model, as one of the generalized linear regression models, is widely used in counting data analysis. With the rapid development of computer technology, more and more variables are obtained and stored, leading to increasingly complex models. In this paper, we consider the sparsity constrained Poisson regression model with L₀ penalty, and apply the Greedy Projected Gradient Newton(GPGN) algorithm to estimate the parameters. The effectiveness of the algorithm is demonstrated through simulation studies on the synthetic dataset, and Poisson regression is applied to the modeling analysis of the prediction floors based on Wi-Fi signals. This verifies the GPGN algorithm performs well in Poisson regression sparsity-constrained optimization.