Machine learning-based prediction of integrated water vapor using meteorological data

Lim, Bing Sheng (2025) Machine learning-based prediction of integrated water vapor using meteorological data. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)

Text (Full Text) Machine learning-based prediction of integrated water vapor using meteorological data.pdf - Submitted Version Download (3MB)

Abstract

This study investigates the use of machine learning models, including Feed-Forward Neural Networks (FFNN), Bagged Trees, and Fine Gaussian Support Vector Machines (SVM), to predict Integrated Water Vapor (IWV) from meteorological data. The models were trained and tested using data from the UTeM FTKEK weather station during Malaysia's northern monsoon season (23 Oct 2019 to 9 Mar 2020). RTKLIB was used to obtain Zenith Total Delay (ZTD) and derive IWV for output of the machine learning models. While FFNN and Bagged Trees excelled in training, Fine Gaussian SVM showed better generalization. Applying moving average filters improved model accuracy by 13%. Additionally, feature selection and time-lag analysis optimized predictions. The study suggests that predicting IWV at weekly intervals yields better results than minute-by-minute forecasts. Future work should focus on expanding the dataset and refining the models for real-world applications.

Actions (login required)

View Item