Multiclass Classification to Predict the Level of Storm and Damages Using Support Vector Machine

We develop an automated storm level forecasting model using the machine learning with Python. We illustrate our methodology using the Dataset NOAA & Climate Prediction Centre by considering 5000 records. Data pre-processing is performed on the data in a dataset to maintain quality data. This process includes cleaning, transformation, and reduction of data. Finally, sampling is done to represent the entire dataset. The sample is selected in order to reduce the complexity and to increase the performance. The dataset splits into trained and testing data (75% and 25%). Model fitting is done next on resulted data using the SVM and storm level is predicted using the parameters like magnitude, length, and width. In this paper, SVM model is compared with KNN and it demonstrates an improvement on the ability to predict the level of storm and damage with the parameters like precision, recall, accuracy, and Fl-score.