Mutual Information Ratio-Based Approach for Rainfall Prediction with Multicollinearity

Multicollinearity is the presence of exact correlation or strongly correlated relationships among the explanatory variables. By analyzing Variance Inflation Factor (VIF), multicollinearity is proven to exist in data analysis of weather forecasting. Mutual Information Ratio (MIR) is an algorithm which has applied in Myanmar weather dataset to address and reduce the effect of multicollinearity existing in reliability analysis. This study has adopted MIR that uses mutual information to maximize the joint dependency with rainfall to identify the most relevant and the least redundant set of factors. This study examines the relationship between rainfall and six explanatory variables, such as maximum temperature, minimum temperature, dry bulb temperature at 9:30am, wet bulb temperature at 9:30am, dew point temperature at 9:30am and humidity at 9:30am. The data were acquired from Myanmar Meteorological Data from Meiktila weather station, as categorical variables. The result given the number of factors required to explain rainfall prediction by reducing multicollinearity problem. Hence, MIR can be recommended as the appropriate method for complex data which is applicable in the domain of rainfall prediction for weather data.