Snowfall-Prediction-using-Machine-Learning
In this research, Machine learning algorithms like Long-Short Term Model (LSTM), Decision tree, Random Forest and XG Boost were used as a classifier to improve the accuracy of Snowfall prediction for the region of Boston. The geographical parameters like Humidity, Temperature, Wind-speed, Precipitation, Sea-level, Dew-point and Visibility were used as independent variables. Before the modeling phase, Data lagging was performed for 2 step followed by Exploratory Data Analysis was using techniques like Multiple Linear Regression, Correlation Plot and variable importance plot. Feature Selection was also executed using Logistic Regression and Boruta algorithm. Experimental evaluations resulted in the highest accuracy shown by LSTM with an accuracy of 89.98%. In terms of sensitivity, Random Forest outperformed other classifier models. Whereas, Decision tree and XG Boost resulted well in the overall performance of prediction with respect to other evaluation metrics. The results of this research added to the contribution of the knowledge in weather prediction in the domain of Snowfall for the machine learning industry.
