Least Square Support Vector Regression Models for Water Quality Index Prediction

Orhan N., Çıtakoğlu H., Öner A. A.

7th International Azerbaijan Congress on Life, Engineering, Mathematical, and Applied Sciences, Baku, Azerbaycan, 25 - 27 Haziran 2024, cilt.1, sa.1, ss.163

  • Yayın Türü: Bildiri / Özet Bildiri
  • Cilt numarası: 1
  • Basıldığı Şehir: Baku
  • Basıldığı Ülke: Azerbaycan
  • Sayfa Sayıları: ss.163
  • Erciyes Üniversitesi Adresli: Evet

Özet

The Water Quality Index (WQI) serves as a metric for assessing water quality at specific locations over defined time periods. A high value on the index signifies that the water is unsuitable for drinking and does not meet the required standards for designated uses due to inadequate quality. This study explores the suitability of two distinct soft computing methodologies: Least Square Support Vector Regression (LSSVR) and Support Vector Regression (SVR) for estimated WQI. The determination of WQI involves highly intricate calculation procedure that encompasses several individual water quality variables, including pH, total Cyanide, Sulfate, Nitrate, Sodium, and Free Chlorine, each contributing to the overall magnitude of water quality assessment. The models were applied to monthly data collected from Kayseri Water and Sewerage Admınıstratıon. The models were compared with each other using mean absolute error, relative root mean squared error, mean absolute percentage error, overall index of model performance, Kling-Gupta efficiency, Wilmott’s refined index, Nash–Sutcliffe efficiency, and determination coefficient. According to the comparison, LSSVR provided better accuracy than the SVR. The results indicated that both methods yielded satisfactory outcomes in modeling WQI. At the conclusion of this study, the efficacy of the recommended methods for comparison was validated using the Kruskal–Wallis test. The findings of this research present a validated soft computing model for determining WQI, offering a viable alternative to conventional procedures that are often time-consuming, costly, labor-intensive, and occasionally prone to computational errors. 

Keywords: Water quality index, LSSVR, SVR, Kayseri Water and Sewerage Administration. The authors thank the Scientific Research Projects Unit of Erciyes University under the contract numbers FYL-2024-13494 for funding.