MODELLING ANALYSIS FOR ACCURATE TROPICAL WEATHER FORECASTING

Calvin, Wong Qin Jie (2023) MODELLING ANALYSIS FOR ACCURATE TROPICAL WEATHER FORECASTING. [Final Year Project Report] (Unpublished)

	PDF Calvin Wong 24pgs.pdf Download (15B)
	PDF (Please get the password by email to repository@unimas.my, or call ext: 3914/ 3942/ 3933) Calvin Wong ft.pdf Restricted to Registered users only Download (5MB)

Abstract

One of the most difficult aspects of weather forecasting is tropical weather forecasting. Rainfall prediction is to forecast the rainfall at a specific location with the use of science and technology. Predicting rain accurately and in a timely manner is crucial for many reasons, including agricultural work, flood situations, transportation activities, ongoing construction projects, airline operations and so on. Over the last decade, numerous research have attempted to enhance rainfall prediction accuracy by improving and merging data mining methodologies. In this research, the Random Forest, K-Nearest Neighbors, Support Vector Machines, XGBoost and Naïve Bayes algorithm are used to forecast rainfall. Five years of historical weather data (2018 - 2022) for Kuching, Sarawak are utilised for forecasting. The weather attributes in the dataset were obtained from the Kuching airport weather station. The prediction model was trained on training data and then assessed for accuracy on test data. The data set for this project is segmented into training and testing data proportions with 75% for training and 25% for testing, or 85% for training and 15% for testing, respectively. The Random Forest, K-Nearest Neighbors, Support Vector Machines, XGBoost and Naïve Bayes algorithm is proposed to validate the model for rainfall prediction, which is proven to operate well with excellent accuracy in previous researches. In summary, the Naive Bayes approach for rainfall prediction demonstrated impressive accuracy, precision, recall, and F1-score. It also exhibited remarkable computational efficiency, delivering reliable predictions within a short computation time.

Actions (For repository members only: login required)

View Item