APPLICATION OF GIS AND STATISTICAL METHODS TO SELECT OPTIMUM MODEL FOR MALARIA SUSCEPTIBILITY ZONATION: A CASE STUDY

Abstract

The representation and analysis of maps of malaria-incidence data is a basic tool in the analysis of regional variation in public health. An attempt has been made for Varanasi district, India to develop malaria susceptibility model using different statistical methods, by which malaria prone zones could be predicted using five classes of malaria susceptibility and comparison of statistical methods to select optimum model for malaria susceptibility zone and verification of the susceptibility zone by area under curve (AUC) though Remote Sensing data and GIS. Multiple linear regression, Information value and heuristic methods are applied for malaria disease occurrence. Using the causal factors and indicators, malaria susceptibility index (MSI) and malaria susceptibility zones (MSZ) are developed. Malaria density ratio (Qs) is used to calculate optimum model for malaria susceptibility index and malaria susceptibility zones. The verification method is performed by comparison of existing malaria data and malaria analysis results by area under curve (AUC). It is found that the information value method having Qs =3.96 has been selected as an optimum model for malaria susceptibility zonation in the study area, whereas Qs value for Heuristics method and Multiple linear regression method are 1.67 and 1.43 respectively. Verification results show that in the information value case, the area under curve (AUC) is 0.696 and the prediction accuracy is 69.60%. In the heuristic and multiple linear regression case, the AUC is 0.603 and 0.484 and the prediction accuracy is 60.30% and 48.40% respectively.