Explainable Machine Learning Framework for Detecting Lumpy Skin Disease with Environmental and Climate Factors

Lumpy Skin Disease (LSD) is a viral illness af- fecting cattle, marked by rapid transmission and significant economic losses. Management challenges and diagnosis issues make LSD a major health and financial risk for the livestock sector. However, this study presents a comprehensive comparative analysis of eleven machine learning algorithms specifically aimed at binary classification within an environmental context. The evaluated algorithms comprise traditional classifiers, ensemble techniques, and advanced boosting methods, including support vector machine, k-nearest neighbors, random forest, decision tree, CatBoost, XGBoost, AdaBoost, linear classification, as well as stacking and voting classifiers. Evaluation was based on performance metrics like accuracy, precision, recall, and F1- score, with CatBoost achieving the highest results (98% accu- racy, 91% precision, 91% recall, and 91% F1-score). Typically, ensemble methods surpassed standalone models in effectiveness. An assessment of feature importance, using CatBoost and SHAP (SHapley Additive exPlanations), identified cloud cover, diurnal temperature range, and frost days as vital predictors, corrob- orated by historical climate data from 2010 that demonstrated notable predictive relevance. The application of interpretability tools ensured that top-performing models remained transparent and actionable. This research offers a robust, interpretable, and flexible classification framework appropriate for environmental modeling and wider machine learning applications, underscoring the value of algorithm comparison and varied feature utilization for effective predictive modeling.