Data mining techniques for predicting dengue outbreak in geospatial domain using weather parameters for New Delhi, India

Abstract

Dengue is a hazardous disease which poses a critical threat to the population of Delhi, India. These cases are steadily reported during and post-monsoon season indicating its correlation with weather parameters. Establishing this relation will help understand the spread of dengue and will allow decision makers take precautionary steps beforehand. Our study explains the adopted multi-regression and Naïve Bayes approach to model the relation between dengue cases and weather parameters, i.e. maximum temperature, rainfall and relative humidity. Both these models have served a great deal in modelling this relationship which has enabled us to forecast a probable dengue outbreak. Our results have shown that sudden and high rainfall accompanied with 30-35°C temperature and high relative humidity contributes to a highly vulnerable weather for the spread of dengue. Also, we have proposed a new application of spherical k-means clustering algorithm to identify zones with similar transmission pattern which gives insight into the distribution of dengue incidences in Delhi. Results show that Central, Civil Lines, Rohini, South and West zones have the highest odds of dengue occurrences.