Exploring the Effectiveness of GRU Model with SIELU Activation Function for Water Quality Classification

Abstract

Accurate water quality classification is crucial for effective environmental monitoring and resource management. This research investigates the efficacy of the Gaussian Error Linear Unit with Sigmoid (SIELU) activation function in a Gated Recurrent Unit (GRU) model for improved water quality classification accuracy. Specifically focusing on Hong Kong rivers, this study justifies the dataset selection due to its consistent historical data, minimal missing values, and diverse parameters encompassing classifications of good, fair, and poor. By leveraging the unique properties of the SIELU activation function, the proposed SIELU-GRU model aims to enhance the GRU’s performance and generalization capabilities in the context of Hong Kong water quality classification. The research includes rigorous experimentation and analysis, comparing the proposed model with the existing approach, the Gaussian Error Linear Unit (GELU) activation function. Results highlight the superior accuracy achieved by a small margin by the model that includes GELU activation function, indicating its potential for improving environmental monitoring systems, aiding decision-making processes, and facilitating resource management, compared to the model that includes SIELU activation function. This study contributes to advancements in water quality classification methodologies, ultimately benefiting the sustainability and well-being of water ecosystems in the world.