Performance, evaluation and prediction of weather and cyclone categorization using various algorithms

The most endangered regions of cyclones formed in the world are Indian sub-continent. The coastal line present in this region is about a total of 7516 km that includes 132 km in Lakshadweep, 5400 km main land, and Andaman and Nicobar Islands includes about 1900 km. The total region consists nearly 10% of tropical cyclones formed around the world. The analyses of data between the years 1980 and 2000 have shown that an average of 370 million people is affected by the cyclone in India. The formation of cyclone is about 30% in Bay of Bengal and 25% in Arabian Sea during the pre-monsoon season. Huge number of death occurs during the cyclone and heavily made losses to the public and private properties. For this reason, cyclone severity prediction is more important and highly crucial. In the second phase of this research, XGBoost (eXtreme Gradient Boosting) algorithm is applied for predicting the formation and severity of cyclones in Bay of Bengal and its performance is analyzed with SVM and RVM classifiers. Due to high technological advancement and development, loss of damages and many lives are saved by predicting the cyclones well in advance like Gaja and Fani by IMD. In the final stage of the research, a hybrid approach by combining genetic algorithm and XGBoost algorithm (GA-XGBoost) is proposed for predicting the tropical cyclone (TC) severity in Bay of Bengal (BoB) obtained from Indian Meteorological Department (IMD). Furthermore, the results of hybrid model for the TC data is proven to be better when compared with the existing algorithms.