Multiple Linear Regression Models on Interval-valued Dengue Data with Interval-valued Climatic Variables

Reported dengue fever cases are increasing day by day in the world as well as in Sri Lanka. Model, Prediction and Control are three major parts of the process of analysis of the dengue incidence which leads to reduce the burden of the dengue. There is an increasing trend in the applications and developments in interval-valued data analysis over recent years. Particularly, under regressions there have being developed various techniques to handle interval-valued dependent and independent variables. Representation of data as intervals is very much useful to capture uncertainty and missing details associated with variables. Further, the predictions in intervals suit well when the situations of exact forecasts may not necessary. In this study interval-valued dengue data with interval-valued minimum temperature, maximum temperature and rainfall from 2009 to 2015 in the Colombo district, Sri Lanka were model using three interval valued regression procedures, namely, Center Method (CM), Center and Range Method (CRM) and Constrained Center and Range Method (CCRM). Predicted dengue cases in a range is particularly important because actions taking towards controlling the dengue do not depend on exact number but on magnitude of the values represent in the interval. Data in the year 2016 used for the validation of the models which is developed under three methods. Root of the mean square error, coefficient of determination as well as square root of variance of the models were used to select the best procedure to predict dengue cases. Among the three regression procedures both CRM and CCRM perform well in predicting monthly dengue cases in Colombo.