Detecting water contaminants using an electrochemical sensor based on improved Transformer Encoder with Genetic Algorithm

Detecting water contaminants is important to mitigate environmental and health problems. Currently, traditional techniques for detecting water contaminants are costly and time-consuming. Thus, the development of a cost-effective, rapid, and AI-based alternative is needed. This paper proposes a novel integrated and systematic framework for detecting water contaminants from electrochemical sensor data using an improved Transformer Encoder optimized with a Genetic Algorithm (TE-GA). The electrochemical sensor is employed to capture voltammetric signals from water contaminant samples. Then, the Transformer Encoder processes these signals to capture patterns for identifying contaminants. A Genetic Algorithm (GA) is applied to optimize the hyperparameters of the Transformer Encoder. Furthermore, this study compares the effectiveness of GA in improving Transformer Encoder performance with other optimization methods, including Grid Search, Random Search, Particle Swarm Optimization, Bayesian Optimization, and Tree-structured Parzen Estimator. The Technique for Order Preference by Similarity to Ideal Solution ranking confirmed that the TE-GA achieved the optimal balance, delivering the highest accuracy (99.17%) while completing the optimization process in 965.78 s, thereby reducing the computation time by up to 57% compared to other computationally heavy methods.