Developing the real-time water quality program using machine learning and API.

The importance of water quality monitoring (e.g., TOC, DO, Chl-a, TN, and TP) is increasing in part of agriculture, water treatment, and policy decision. As the computing power has been increased, we could develop the real-time water quality system. Our system can forecast the water quality after 2 days from now. To simulate the water quality of ND river, the random forest (RF) and artificial neural network (ANN) were adopted. Furthermore, the program provides a user-friendly system using a graphic user interface (GUI). Our prediction program consists of 3 major phases. Phase 1 utilizes an application programming interface (API) to load the data from national institutes (NI). Phase 2 is the simulation of flowrate of ND River. Phase 3 simulates the water quality using machine learning. RF models produced R² values of 0.46, 0.8, 0.59, 0.46, 0.67 for chl-a, DO, TN, TOC, and TP respectively while ANN models resulted in R² values of 0.22, 0.72, 0.53, 0.35, 0.63. Overall, DO shows the most accurate result while TN and TP showed reasonable simulation results, by showing over 0.5 of R². Our study demonstrates that API service with machine learning is useful for simulating real-time water quality.