Analysis and prediction of water quality using deep learning and auto deep learning techniques

Area of Science:

• Environmental Science
• Computer Science
• Data Science

Background:

• Natural water bodies face significant threats from untreated industrial effluents and sewage.
• Ensuring safe water for consumption necessitates rigorous water quality inspection.
• Traditional water quality testing methods are often cumbersome and time-consuming.

Purpose of the Study:

• To explore the suitability of Automated Deep Learning (AutoDL) for Water Quality Assessment.
• To compare the performance of AutoDL against conventional deep learning models.
• To evaluate the efficiency and ease of use of AutoDL in water quality analysis.

Main Methods:

• Utilized AutoDL to automate deep learning pipeline creation for water quality prediction.
• Compared AutoDL performance with conventional deep learning models using Water Quality Index.
• Assessed accuracy for both binary and multiclass water quality classification tasks.

Main Results:

• Conventional deep learning models achieved slightly higher accuracy (1.8% for binary, 1% for multiclass) than AutoDL.
• Conventional models reached ~98%-99% accuracy, while AutoDL achieved ~96%-98%.
• AutoDL demonstrated superior efficiency by simplifying model selection and reducing manual intervention.

Conclusions:

• AutoDL shows promise for streamlining water quality assessment despite a minor accuracy gap.
• The ease of use and efficiency gains make AutoDL a valuable tool for future water quality monitoring.
• Further development of AutoDL could bridge the accuracy gap and enhance its adoption in environmental monitoring.