- Title
- A hybrid optimisation model for water quality prediction using Naive Bayes and SMO algorithms:: The bagging technique
- Creator
- Mcedani, Lazola Asadumodwa
- Subject
- Water quality
- Subject
- Learning classifier systems
- Subject
- Time-series analysis--Data processing
- Date Issued
- 2024-03
- Date
- 2024-03
- Type
- Master's theses
- Type
- text
- Identifier
- http://hdl.handle.net/10353/29911
- Identifier
- vital:79189
- Description
- This study comprehensively evaluates machine learning classifiers to predict water quality using time-series data. The objective is to identify the most effective classifiers, assess the influence of reconfigurations on their performance, and construct a hybrid model using the superior classifiers. Additionally, the study seeks to ascertain the most suitable ensemble technique. The methodology incorporates the knowledge discovery in databases KDD process and scrutinises 56 classifiers available in WEKA. The investigation reveals support vector machine optimisation SMO and Naive Bayes as the leading classifiers, with 74.37 percent and 70.01 percent prediction accuracies, respectively. Performance enhancements were observed following reconfigurations, with the refined SMO model achieving an accuracy of 79.9827 percent on a novel unseen dataset and the adjusted Naive Bayes model reaching 78.3784 percent. The hybrid model, which amalgamates Naive Bayes and SMO, exhibited improved accuracy without compromising efficiency. Bagging was identified as the most influential ensemble technique, delivering the highest accuracy and overall model efficacy. Bagging achieves a prediction accuracy of 86.4865 percent, surpassing the best-performing base model. It also demonstrates a Kappa statistic of 0.7628, indicating substantial agreement with the ground truth, along with an impressive F-measure of 0.859 and other notable metrics, highlighting its robust performance across various evaluation criteria. This study contributes to developing water quality prediction models, providing valuable insights for researchers and practitioners, and enabling more informed decision-making in environmental management, public health, and resource allocation. The research advances algorithm design, optimises techniques, and pioneers novel hybrid models while offering a robust evaluation framework to compare the performance of various classifiers and ensemble techniques.
- Description
- Thesis (MSci) -- Faculty of Science and Agriculture, 2024
- Format
- computer
- Format
- online resource
- Format
- application/pdf
- Format
- 1 online resource (xiv,115 leaves)
- Format
- Publisher
- University of Fort Hare
- Publisher
- Faculty of Science and Agriculture
- Language
- English
- Rights
- rights holder
- Rights
- All Rights Reserved
- Rights
- Open Access
- Hits: 8
- Visitors: 6
- Downloads: 2
Collections
UFH Department of Computer Science
| Thumbnail | File | Description | Size | Format | |||
|---|---|---|---|---|---|---|---|
| View Details Download | SOURCE1 | Dissertation - Final Copy to Library - Mcedani Lazola Asadumodwa (201301207).pdf | 41 MB | Adobe Acrobat PDF | View Details Download |