Machine Learning Combining Expert Features to Differentiate Infection Types Based on Physiological and Biochemical Indicators of Malayan Pangolin

This research retrospectively analyzed blood parameters of rescued Malay pangolins from 2018 to 2021 to differentiate between diseased individuals and those with pathogenic infections, aiming to enhance clinical diagnosis and treatment strategies. We employed clustering, machine learning, ensemble learning, and expert mechanisms for classification and training. Under supervised conditions, we combined this model with ensemble learning to distinguish healthy from diseased samples, achieving 73.08% accuracy. By integrating expert mechanisms for infection data, we identified bacterial, viral, and parasitic infections from the diseased dataset with 87.5% accuracy, a 14.42% improvement. The naive Bayes and random forest models stood out in performance.