Correlation based Feature Selection and Hybrid Machine Learning Approach for Forecasting Disease Outbreaks

Abstract

According to WHO, Dengue is a viral infection transmitted to humans through the bite of infected mosquitoes i.e., Aedes aegypti mosquitoes. There is currently no known cure for dengue or severe dengue. Artificial Intelligence (AI) in the form of Machine Learning (ML) allows software programs to predict outcomes more correctly without explicit instructions. Machine learning algorithms use historical data as input to forecast new output values. The aim of this study is to identify, evaluate and interpret suitable hybrid algorithms/approaches relevant to the application of machine learning in limiting the spread of deadly disease outbreaks. It focuses on finding a way of predicting the next dengue fever local epidemic by comparing the bench mark approaches available until now. For this the study proposes the use of XGBoost coupled with Moving Average Rolling Features in order to learn the long-term temporal relations in the features to get accurate predictions. The dataset used for evaluating the proposed approach contains number of cases in the two locations: San Juan and Iquitos and it includes information on temperature, precipitation, humidity, vegetation, and what time of the year the data was obtained. A correlation analysis-based feature selection along with Moving Average Rolling Features has been used for getting more precise data implemented with ML approach resulting in MS E 11.37 in San Juan and MSE 6.37 in Iquitos.