Estimasi BOD, COD, dan TSS pada Air Sungai Menggunakan Data Sensor pH, DO, dan Temperatur Berbasis Machine Learning

Natamenggala, Reva Fahrian (2025) Estimasi BOD, COD, dan TSS pada Air Sungai Menggunakan Data Sensor pH, DO, dan Temperatur Berbasis Machine Learning. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5009201147-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Kualitas air sungai, seperti Kali Mas di Surabaya, semakin menurun akibat aktivitas manusia dan industri, dengan parameter seperti DO dan BOD seringkali di luar batas baku mutu. Metode konvensional pengujian kualitas air di laboratorium memakan waktu dan biaya. Oleh karena itu, machine learning menawarkan solusi yang efisien untuk prediksi real-time. Dalam penelitian ini, digunakan enam algoritma regresi: Linear Regression, Random Forest, XGBoost, K-Nearest Neighbors, Support Vector Regression (SVR), dan MLP Regression. Data pengukuran di Sungai Kalimas Surabaya diperoleh dari Dinas Lingkungan Hidup Kota Surabaya, mencakup parameter temperature, pH, DO, BOD, COD, dan TSS. Hasil pra-pemrosesan data menunjukkan penghapusan outlier berhasil menstabilkan distribusi data. Dari evaluasi model, K-Nearest Neighbors menunjukkan akurasi tertinggi untuk prediksi BOD (78.45%), sedangkan Linear Regression terbaik untuk COD (71.15%), dan Support Vector Regression untuk TSS (37.15%). Meskipun demikian, akurasi untuk TSS secara umum lebih rendah dibandingkan BOD dan COD di semua model. Korelasi antara fitur input (pH, DO, Temperatur) dengan target menunjukkan hubungan yang umumnya lemah, kecuali korelasi positif kuat antara BOD dan COD (0.69). Penelitian ini menunjukkan bahwa model machine learning dapat menjadi alternatif yang lebih cepat dan efisien untuk estimasi parameter kualitas air dengan input sensor yang terbatas.
=====================================================================================================================================
The water quality of rivers, such as the Kali Mas in Surabaya, is declining due to human and industrial activities, with parameters like DO (Dissolved Oxygen) and BOD (Biochemical Oxygen Demand) often exceeding quality standards. Conventional laboratory methods for water quality testing are time-consuming and costly. Therefore, machine learning offers an efficient solution for real-time prediction. In this study, six regression algorithms were used: Linear Regression, Random Forest, XGBoost, K-Nearest Neighbors, Support Vector Regression (SVR), and MLP Regression. Measurement data from the Kali Mas River in Surabaya was obtained from the Surabaya City Environmental Agency, covering parameters such as temperature, pH, DO, BOD, COD, and TSS. Data pre-processing results showed that outlier removal successfully stabilized data distribution. From the model evaluation, K-Nearest Neighbors showed the highest accuracy for iprediction (78.45%), while Linear Regression was best for COD (71.15%), and Support Vector Regression for TSS (37.15%). Nevertheless, the accuracy for TSS was generally lower compared to BOD and COD across all models. The correlation between input features (pH, DO, Temperature) and targets generally showed a weak relationship, except for a strong positive correlation between BOD and COD (0.69). This research indicates that machine learning models can be a faster and more efficient alternative for estimating water quality parameters with limited sensor input.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	BOD, pH, Machine Learning, Kualitas air BOD, pH, Machine Learning, Water quality
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK7882.P3 Pattern recognition systems
Divisions:	Faculty of Industrial Technology and Systems Engineering (INDSYS) > Physics Engineering > 30201-(S1) Undergraduate Thesis
Depositing User:	Reva Fahrian Natamenggala
Date Deposited:	07 Aug 2025 03:23
Last Modified:	07 Aug 2025 03:23
URI:	http://repository.its.ac.id/id/eprint/122282

Actions (login required)

View Item