IoT and Machine Learning Based System for Predicting Time to pH or TDS Thresholds in Refill Drinking Water

Hani, Aminah Qalbu (2026) IoT and Machine Learning Based System for Predicting Time to pH or TDS Thresholds in Refill Drinking Water. S1 thesis, Universitas Andalas.

	Text (Cover dan Abstrak) 01. Cover dan Abstrak .pdf - Published Version Download (169kB)
	Text (BAB I) 02. BAB I .pdf - Published Version Download (96kB)
	Text (BAB V) 03. BAB V .pdf - Published Version Download (133kB)
	Text (Daftar Pustaka) 04. Daftar Pustaka.pdf - Published Version Download (181kB)
	Text (Skripsi Fulltext) 05. Skripsi Fulltext.pdf - Published Version Restricted to Repository staff only Download (1MB) | Request a copy

Abstract

Drinking water is a basic necessity that plays an essential role in maintaining human health and supporting daily activities. Refill drinking water has become one of the preferred choices because it is available at a relatively affordable cost. Monitoring drinking water parameters is important to help users obtain information regarding changes in the condition of the water used daily. This study aims to develop a prototype of an IoT and machine learning based system for predicting the remaining time until TDS reach ≥ 300 ppm or pH falls outside the permissible range of 6.5 to 8.5 in refill drinking water. The IoT technology in this system allows sensor measurement data and prediction results from the ESP32 to a Cloud Firestore database to be displayed on a mobile application. The system is equipped with a DS18B20 temperature sensor, an SEN0244 TDS sensor, and a 4502C pH sensor. The system uses three machine learning models including Random Forest Regressor, XGB Regressor and SVR to predict time. Based on the evaluation results, the Random Forest Regressor was selected for implementation in the system, achieving a MAE of 0.3388 hours, a RMSE of 0.9675 hours, and an R2 of 0.9990. The system recorded 26,480 monitoring data points together with their corresponding prediction results. TDS was the first parameter to exceed the permissible limit, reaching 302.01 ppm, whereas the pH remained within the permissible range of 6.5 to 8.5.

Actions (login required)

View Item