A unified LSTM autoencoder–random forest AI-IoT framework for real-time water quality monitoring, anomaly detection, and water quality index prediction

Real-time water quality monitoring is essential for protecting aquatic ecosystems and public health; however, existing intelligent monitoring systems typically address either anomaly detection or Water Quality Index (WQI) prediction independently, limiting their effectiveness for continuous environmental surveillance. This study proposes a unified Artificial Intelligence–Internet of Things (AI-IoT) framework that integrates a Long Short-Term Memory Autoencoder (LSTM-AE) for unsupervised temporal anomaly detection with a Random Forest regressor for WQI prediction within a coordinated decision-support architecture. The novelty of the proposed framework lies in the architectural integration of complementary learning models to simultaneously detect abnormal water quality conditions and estimate WQI in real time, rather than introducing new machine learning algorithms. A synthetic multivariate IoT water quality dataset comprising 10,000 observations was developed to emulate realistic environmental monitoring conditions, including temporal dependencies, sensor noise, drift, seasonal variations, and contamination events. Experimental evaluation demonstrated that the proposed LSTM Autoencoder achieved a ROC-AUC of 0.9763, outperforming Isolation Forest ( 0.8914 ) and One-Class SVM (0.8732) for anomaly detection. For WQI prediction, the Random Forest regressor achieved the lowest prediction error with an MSE of 2.3573, surpassing Gradient Boosting (2.6841), Sequence-to-One Recurrent Neural Network (2.9137), and Support Vector Regression (3.1048 ). These results demonstrate that integrating temporal anomaly detection with ensemble-based regression provides a robust and scalable solution for intelligent water quality monitoring. Although validated using a reproducible synthetic benchmark, the framework establishes a practical foundation for AI-enabled environmental monitoring, with future work focusing on validation using real-world IoT water quality datasets.