An Explainable Machine Learning Model For Lumpy Skin Disease Occurrence Detection
Anuj Kumar Jain, Raj Gaurang Tiwari, Neha Ujjwal, Anshbir Singh
Abstract
Humans depend greatly on animal companions. For the most part, we humans rely on animal products such as milk, curd, honey, etc. Therefore, it is humanity’s highest duty to ensure their well-being. In today’s world, cattle and water buffalo are experiencing an epidemic of a condition known as a lumpy skin disease (LSD). Skin nodules are the hallmark of this contagious, eruptive, and sometimes fatal illness. In this research we aim to make some kind of prognosis about whether or not the cattle in a certain area are now, or may perhaps in the future, be infected with this illness. We have used many machine learning algorithms to make predictions about lumpy skin conditions and compared their performance. There are a total of 18603 cases and 16 characteristics in the dataset, with target columns that may take on either 0 (indicating the absence of lumpy illness) or 1 (indicating its presence). We determined that the RandomForest method had the greatest accuracy (97.7 percent) of all the algorithms we tested. Since Random Forest proves to be the most effective for the chosen dataset, we additionally employ SHAP (SHapley Additive exPlanations) visualizations to learn more about its classification accuracy.