Disease Prediction Model to Assess the Impact of Changes in Precipitation Level on the Risk of Anthrax Infectiousness among the Livestock Hosts in Karnataka, India
Anthrax is a one of the zoonotic diseases existing in India. Early detection of anthrax outbreaks is crucial for minimizing anthrax morbidity and death, as well as the risk of anthrax transmission in the population. Objective of the present research is to develop a disease prediction model by employing Machine-Learning techniques to assess the risk of anthrax analogous to the impact of changes in precipitation level that can benefit as an early warning system for detecting future anthrax outbreaks among livestock across Karnataka. By considering the disease incidence data during 2000 to 2019, livestock population data and the ecological parameters, the machine learning model was successful in identifying the next outbreak susceptible areas and the parameters that contribute significantly to the disease outbreak. Machine learning model was developed by R statistical software version 3.1.3 using different data mining regression and classification models viz., GLM, GAM, MARS, FDA, CT, SVM, NB, ADA, RF, GBM and ANN. Disease incidence data was collected from Department of animal husbandry, Bengaluru, Karnataka. Disease incidence data was divided in two groups based on average annual precipitation above and below normal (1151mm) for the risk assessment and study the impact of changes in precipitation level. Data with average annual-precipitation above normal was predicted with high risk in the north, northern east and the state's southern region. Whereas data with average annual-precipitation below normal was predicted with high risk in south, northern east and the state's central region. Cohen's Kappa, ROC curve, True Skill Statistics (TSS), and ACCURACY was used to assess the models' performance. Further, this model can be intensified and validated using the anthrax outbreak data available at national level which will be useful for policymakers to formulate control strategies.