|
Approach of Deep Learning Model Based Multi-Layer Feed-Forward Artificial Neural Network with Backpropagation Algorithm for Water Quality Prediction |
|---|---|
| รหัสดีโอไอ | |
| Creator | 1. Chalisa Veesommai Sillberg 2. Thitima Rungratanaubon |
| Title | Approach of Deep Learning Model Based Multi-Layer Feed-Forward Artificial Neural Network with Backpropagation Algorithm for Water Quality Prediction |
| Publisher | Thai Society of Higher Education Institutes on Environment |
| Publication Year | 2565 |
| Journal Title | EnvironmentAsia |
| Journal Vol. | 15 |
| Journal No. | 1 |
| Page no. | 1-11 |
| Keyword | Environmental and information analysis, Supervised data, Data prediction, River water-quality, Chao Phraya River |
| URL Website | https://tshe.org/main/ea-journal-online |
| Website title | EnvironmentAsia Journal |
| ISSN | 1906-1714 |
| Abstract | Quality of water resources is crucial to several activities in human life and life under water. Water resource quality is a complex dynamic system depending on various attributes with nonlinear and non-equilibrium conditions. In this study, we aim to utilize a deep learning (DL) model based on multi-layer feed-forward artificial neural network with back-propagation algorithm and supervised data for water quality prediction. The DL model predicts the water quality from database information that collected by the Pollution Control Department (PCD), Thailand. The following water quality attributes from 20072019 were analyzed: biological oxygen demand (BOD), dissolved oxygen (DO), fecal coliform bacteria (FCB), total coliform bacteria (TCB), ammonia-nitrogen (NH3-N), and spatial-temporal data for the Chao Phraya River in Thailand. The results indicated that the DL model had a coefficient of determination (R2) of 0.90, with a mean square error (MSE) of 0.20 and a predicted accuracy of 88.96%. The evaluation DL model result was more precise in predicting values in the unsuitableand poor water quality classes, with 100% of the dataset being in those classes. The DL model was able to process the data at circa 18,000 rows per second. These results supported the possibility of using the DL model as a tool to predict the quality of water resources. |
