The Image Classification Method with CNN-XGBoost Model Based on Adaptive Particle Swarm Optimization

CNN is particularly effective in extracting spatial features. However, the single-layer classifier constructed by activation function easily interfered image noise, resulting reduced classification accuracy. To solve problem, advanced ensemble model XGBoost used to overcome deficiency of a single classify further distinguish extracted features, CNN-XGBoost optimized APSO proposed, where optimizes hyper-parameters on overall architecture promote fusion two-stage model. The mainly composed two parts: feature extractor CNN, which automatically extract features from images; applied after convolution. In process parameter optimization, shortcoming that traditional PSO algorithm falls into local optimal, improved guide particles search for optimization space different strategies, improves diversity particle population and prevents becoming trapped optima. results set show proposed gets better classification. Moreover, APSO-XGBoost performs well credit data, indicates has good ability scoring.