Implementing predictive analytics for water treatment process optimization at Sebatik Island

Assan, Mohamad Shahrinas (2026) Implementing predictive analytics for water treatment process optimization at Sebatik Island. [Student Project] (Unpublished)
Abstract

Water treatment operations at Sebatik Island rely on real-time tank water-level monitoring through the Aeris Artificial Intelligence of Things (AIoT) device and satellite connectivity. However, the existing monitoring process provides limited support for predicting future water-level conditions, analyzing environmental influences and comparing alternative operating-hourscenarios. This project aimed to identify the current water treatment process and challenges at Sebatik Island, implement predictive analytics for water treatment process optimization and visualize the prediction results through a dashboard. The study reviewed related concepts of predictive analytics, AIoT based monitoring, water-level prediction and dashboard decision support to establish the project foundation. The CRISP-DM methodology was applied, involving business understanding, data understanding, data preparation, modelling, evaluation and deployment. Historical water-level, operational, time-based and environmental data were prepared for regression modelling. Linear Regression, Random Forest Regression and Decision Tree Regression were developed and evaluated using 5-fold and 10-fold cross-validation. Model performance was assessed using Root Mean Squared Error (RMSE), Mean Absolute Error (MAE) and R-squared (R2). Decision Tree Regression using 5-fold cross-validation achieved the best result, with an RMSE of 0.388, MAE of 0.164, and R2 of 0.948. The selected model was deployed through a Power BI dashboard containing water-level monitoring, prediction and environmental impact analysis, operational scenario comparison, decision support and user guideline pages. The dashboard identified the 20-hour operating scenario as the best overall option because it maintained normal predicted water-level conditions and full time-period coverage while reducing operating duration compared with the 24-hour baseline. Expert evaluation findings showed that the dashboard was functional, learnable, easy to use and satisfactory for operational monitoring. Overall, the project demonstrates that predictive analytics could enhance real-time monitoring by providing scenario-based decision support for water treatment process optimization at Sebatik Island.

Item Details
Edit Item
Edit Item
Downloads & Files
[thumbnail of 149557.pdf]
Text
149557.pdf
Download (100kB)
Location & Physical Holdings
Digital Copy
Physical Location
Item Status
Indexing & Metrics