Real-Time Weather Data Analysis for Disaster Management and Response

Real-time detection of weather can be used to understand and handle disasters. Weather conditions are to be foreseen to tackle the disasters caused by extreme weather conditions. A weather dataset usually holds a lot of missing values that may badly affect the prediction accuracy and further decision-making. In this paper, we will handle the missing data problem by filling the missing values using backward and forward-filling techniques in order to regain completeness in the dataset. These imputation methods ensure that the training and evaluation of models will be well-grounded. In this regard, we have used two machine learning algorithms, namely Random Forest and Decision Tree, to validate the accuracy of our prediction. Each of these techniques has inherent strengths in modelling complicated patterns and non-linear relationships. To this end, we further demonstrate that the data imputation techniques presented here coupled with advanced machine learning algorithms ensure a significant improvement in the reliability of real-time weather detection by comparing the performance of these models.