Short-term Weather Prediction using Historical Data: Evaluating Supervised Machine Learning Techniques

Weather Forecasting or Predictions plays an important role in human life as it impacts agriculture, transportation and natural disasters. Therefore, the development of new techniques in machine learning is continuing to enable better and accurate predictions of weather using historical data such as temperature, dew point temperature, relative humidity, visibility, pressure, and wind speed over the years. The primary goal of my system is to develop a model that will be precise and accurate for short-term predictions. This proposed system will first do data preprocessing which involves dealing with missing or null values. Then the architecture applies several machine learning algorithms such as logistic regression, Decision tree classifier, Random forest classifier, SVC, K neighbor classifiers and Gaussian Naive Bayes. Now each model is evaluated based on the prediction accuracy by MAE and RMSE. After comparison, the proposed system confirms that the random forest model indeed outperformed all other models.