Lumpy Skin Disease Classification on Imbalanced Data: A Comprehensive Approach Using Machine and Ensemble Learning

Lumpy Skin Disease (LSD) is a major problem of veterinary care that requires proper, prompt diagnosis and classification to enable effective treatment. However, medical datasets are often affected by such challenges as class imbalance, missing values, outliers, and high dimensionality, which makes it challenging to create effective diagnostic  tools. This study presents a hybrid machine learning (ML) model which aims to address these typical data complexities. We integrate the Synthetic Minority Over-Sampling Technique (SMOTE) as an effective method of dealing with data imbalance with ensemble learning, Bagging in particular, to improve the precision of the classification. In our research, three well-known ML algorithms Decision Trees (DT), Logistic Regression (LR) and Naive Bayes (NB) are rigorously tested together with our Bagging model. The evaluation of the performance was performed on a wide range of measures: accuracy, precision, recall, F-score, and Matthews Correlation Coefficient (MCC). The findings are conclusive in the sense that the Bagging ensemble has been shown to be better than the rest of the models in the sense that it had the highest accuracy of 89.5% and the highest precision and recall rates as well. Moreover, SMOTE was very useful in reducing the bias in the dataset thus making model training more accurate. These results highlight the significant performance of SMOTE and ensemble learning combination towards producing very reliable and accurate diagnostic tools. Such an original method can bring a major breakthrough in LSD treatment and has an enormous potential to enhance the diagnosis of other multifactorial disorders in veterinary medicine.