GBMPSO: Hybrid Gradient Boosting Machines with Particle Swarm Optimization in Cell Segmentation Data

This paper describes a hybrid computing model, named GBMPSO, that enhances the performance of gradient boosting machines (GBMs) whose hyperparameters are optimized by particle swarm optimization algorithm. GBMs are tree-based ensemble algorithms that are used for various machine learning tasks. Particle swarm optimization is a stochastic, metaheuristic swarm-based optimization algorithm. The performance of GBMPSO is tested on a cell segmentation dataset. Our experiment shows the outperformance of GBMPSO over a random search-based GBM and Xgboost, highlighting the potential superiority of GBMPSO over other models and its application to classification problems of other problem domains.