A Factual Flash Flood Evaluation using SVM and K-NN

In recent decades diversified strategies have been applied for the prediction of flash floods more accurately as it is one of the most dangerous natural disaster in the world. Flash floods are usually very rapid and sudden therefore forecasting the exact time of flood and venue can be acknowledged as a complex task. In the recent era Machine learning and Artificial Intelligence based predictive models have been developed for the robust and vigorous forecasting of floods. Data set from the different appropriate sensors can be collected and processed through any predictive algorithm. Prediction can be based on the historical data as well therefore in this research paper the data has been collected from the Pakistan Meteorological Department (PMD) for the duration of 2016 to 2017. The data can be categorized into three levels of danger. Level 3 is the lowest level, Level 2 was set as the moderate level and Level 3 as a dangerous level. These levels must be announced and communicated to the people of the vicinity of the affected area for the evacuation routes. Robust classification of these levels are mandatory in order to classify the true event therefore very popular classifiers named as support vector machine and K-NN classifiers have been applied on the data set for the classification of levels accurately and precisely. Simulated results on the filtered data proved that Support vector machine (SVM) performed better classification compared to the K-NN algorithm in terms of efficiency, confusion matrix, precision, recall and fl-score and prediction speed.