Portfolio
Flood Prediction: ML Analysis
-
Client Personal
-
Category AI Engineering
-
Date July, 2023
In this group project at the university, our team tackled the challenge of predicting future floods in the UK using a limited dataset of historical flood events. As the project leader, I guided our efforts in exploring and comparing various predictive models, ranging from traditional machine learning techniques to more advanced deep learning approaches. Our primary objective was to identify which model could achieve the highest accuracy in forecasting floods, despite the constraints posed by the small volume of training data available.
Our approach involved:
- Conducting comprehensive analysis and preprocessing of the UK flood history data to optimize it for training purposes.
- Experimenting with multiple machine learning models, such as decision trees and support vector machines, to establish a baseline for prediction accuracy.
- Advancing to deep learning models, including convolutional neural networks (CNNs) and recurrent neural networks (RNNs), to explore their potential in capturing complex patterns within the limited data.
- Implementing cross-validation techniques to rigorously evaluate each model's performance and ensure reliability in our predictions.
This project not only demonstrated our team's ability to navigate the challenges of working with sparse datasets but also allowed us to contribute valuable insights into the application of artificial intelligence in environmental science. Through diligent research, experimentation, and leadership, we were able to present a comprehensive comparison of predictive models, highlighting a path forward for enhancing flood prediction efforts with limited data.