A Feature Selection Based on Improved Artificial Hummingbird Algorithm Using Random Opposition-Based Learning for Solving Waste Classification Problem

Recycling tasks are the most effective method for reducing waste generation, protecting environment, and boosting overall national economy. The productivity effectiveness of recycling process strongly dependent on cleanliness precision processed primary sources. However, operations often labor intensive, computer vision deep learning (DL) techniques aid in automatically detecting classifying trash types during chores. Due to dimensional challenge posed by pre-trained CNN networks, scientific community has developed numerous inspired biology, swarm intelligence theory, physics, mathematical rules. This research applies a new meta-heuristic algorithm called artificial hummingbird (AHA) solving classification problem based feature selection. performance AHA is barely satisfactory; it may be stuck optimal local regions or have slow convergence. To overcome these limitations, this paper develops two improved versions AHA-ROBL AHA-OBL. These enhance exploitation stage using random opposition-based (ROBL) (OBL) prevent optima accelerate main purpose apply AHA-OBL select relevant features provided models (VGG19 & ResNet20) recognize classification. TrashNet dataset used verify proposed approaches (the AHA-OBL). suggested methods AHA-OBL) compared with that 12 modern competitive optimizers, namely (AHA), Harris hawks optimizer (HHO), Salp (SSA), aquila (AO), Henry gas solubility (HGSO), particle (PSO), grey wolf (GWO), Archimedes optimization (AOA), manta ray foraging (MRFO), sine cosine (SCA), marine predators (MPA), rescue (SAR). A fair evaluation algorithms’ achieved same dataset. analysis algorithms applied terms different measures. experimental results confirm superiority over other comparative algorithms. produce number selected highest degree precision.