ANNE (Artificial Neural Network for E-coli) is an AI-based predictive model developed to simulate the spatio-temporal distribution of E. coli concentrations in river systems. As shellfish production areas—typically located in shallow coastal waters or embayments—are highly vulnerable to land-based microbial and chemical contaminants, ANNE provides a scientific foundation for the sanitary management of these critical marine environments.

• Prediction of Contribution Rates by Land-based Pollution Sources in the Tongyeong-Jaran Bay Watershed / National Institute of Fisheries Science (2019)

Artificial Neural Network Architecture:

• Input Layer: Incorporates diverse variables including watershed attributes, precipitation, and air temperature.

• Hidden Layer: Dynamically optimizes the model by varying the number of layers (1 to 20) to identify the most effective architecture.

• Optimization: Utilizes Mean Absolute Error (MAE) as the loss function and ReLU (Rectified Linear Unit) as the activation function for robust performance.

• Framework & Tools: Developed using industry-standard deep learning libraries, Keras and TensorFlow.

Automated System Integration:

• Data Acquisition: Automatic crawling and collection of water quality, streamflow, and meteorological data.

• Preprocessing: Automated outlier detection, removal, and interpolation to generate high-quality continuous datasets.

• Autonomous Model Building: Automatically evaluates and selects the optimal number of hidden layers, nodes, and functions to construct the best-performing model.