The implementation of real-time weather forecasting system using Internet of Things and machine learning

Pandya, B ORCID logoORCID: https://orcid.org/0000-0002-9396-1286, Al-Habaibeh, A ORCID logoORCID: https://orcid.org/0000-0002-9867-6011, Rolleston, G, Farnsworth, M and Bansal, D, 2024. The implementation of real-time weather forecasting system using Internet of Things and machine learning. In: Riffat, S, ed., Sustainable Energy Technologies: proceedings of the 21st International Conference on Sustainable Energy Technologies, 12 to 14th August 2024, Shanghai, China. University of Nottingham, pp. 391-400. ISBN 9780853583561

[thumbnail of 2335196_Al-Habaibeh.pdf]
Preview
Text
2335196_Al-Habaibeh.pdf - Published version

Download (5MB) | Preview

Abstract

This research explores the application of advanced technologies to enhance the precision, accessibility, and effectiveness of weather forecasts, presenting benefits to individuals, businesses, and communities. The research utilises data collected from IoT-enabled weather stations, employing appropriate IoT protocols for data acquisition. This data undergoes pre-processing and normalisation to facilitate further analysis. The primary objective is to monitor and predict weather changes, specifically focusing on the probability of rainfall in London. This is achieved through a detailed examination of various meteorological parameters such as temperature, humidity, wind speed, dew point, and wind gusts. Through feature engineering, critical predictors are identified and optimised by eliminating redundant elements, thus refining the model's efficiency. Key features such as temperature, humidity, wind speed, gusts, air pressure, and dew point are analysed alongside temporal variables like time of day, day of the week, and seasonal patterns. The weather conditions are classified into three categories: Cloudy, Fair, and Rain. The dataset spans from 2014 to 2023, with a 70% split for training and 30% reserved for testing. Upon evaluating 17 distinct classifiers, the Support Vector Machine (SVM) classifier emerged as the most effective, demonstrating an 88% recall, 70% precision, and a 78% F1-score. These findings highlight the potential of integrating machine learning with real-time weather monitoring to predict weather patterns accurately.

Item Type: Chapter in book
Description: Paper presented at SET2024: 21st International Conference on Sustainable Energy Technologies 12 to 14th August 2024, Shanghai, China.
Creators: Pandya, B., Al-Habaibeh, A., Rolleston, G., Farnsworth, M. and Bansal, D.
Publisher: University of Nottingham
Date: 12 December 2024
Volume: 4
ISBN: 9780853583561
Identifiers:
Number
Type
2335196
Other
Rights: © 2024 Copyright University of Nottingham & WSSET
Divisions: Schools > School of Architecture, Design and the Built Environment
Record created by: Laura Borcherds
Date Added: 20 Feb 2025 16:11
Last Modified: 24 Feb 2025 12:02
URI: https://irep.ntu.ac.uk/id/eprint/53099

Actions (login required)

Edit View Edit View

Statistics

Views

Views per month over past year

Downloads

Downloads per month over past year