Rivers play a vital role in maintaining ecological balance. Accurate water quality predictions help manage pollution levels, protect aquatic life, and preserve biodiversity. Traditional water quality testing methods are often time-consuming and less accurate. Predictive models using machine learning and deep learning offer more precise and timely assessments, enabling proactive management strategies. This research aimed to create a dynamic system for predicting river water conditions by employing one standalone model and two hybrid models. The study analyzed four key parameters—pH, salinity, turbidity, and water temperature—across three rivers in Hong Kong, namely Kam Tin, Tung Chung, and Lam Tsuen. Data spanning 24 years (2000–2023) was utilized to forecast monthly levels of total dissolved solids (TDS). To analyze the collected data, it was partitioned into two sets: 80% for training and 20% for testing. The data was then utilized in various models, including a convolutional neural network (CNN), a long short-term memory (CNN-LSTM), a gate recurrent unit (CNN-LSTM-GRU), and a deep neural network (DNN). The performance of the models was assessed and compared using three metrics: root mean square error (RMSE), coefficient of determination (R²), and Nash-Sutcliffe (NSE). The findings indicated that the hybrid CNN-LSTM-GRU model outperformed the other two models in terms of accuracy and efficiency. This model has the potential to be applied to other rivers, enhancing the speed and effectiveness of water quality monitoring and control.

The rapid expansion of the digital economy is reshaping the global production and trade system, bringing new opportunities for developing economies seeking to enhance their international competitiveness, while also posing structural challenges. This study focuses on China, a typical emerging economy, and uses provincial panel data from 2015 to 2024 to empirically examine how digital transformation and technological innovation jointly affect foreign trade competitiveness. The core variables are measured as follows: The digitalization level is constructed using principal component analysis (PCA) based on three dimensions: digital infrastructure, digital industrialization, and industrial digitization; technological innovation is proxied by the logarithm of technology market transaction volume. This study employs a fixed-effects model with interaction terms to estimate the independent effects of digitalization and technological innovation and to explore their interaction within the framework of the digital economy. The empirical results show that both digital transformation and technological innovation have a significant positive impact on foreign trade competitiveness. Specifically, a 10-point increase in the digitalization index is associated with an approximately 0.10-unit increase in the trade competitiveness index, and a 1% increase in technological innovation input is associated with an increase of 0.032–0.042 units. However, their interaction coefficient is significantly negative (−0.001, p < 0.01), indicating a substitution effect: an increase in technological innovation investment weakens the marginal contribution of digitalization to export competitiveness, and vice versa. Further heterogeneity analysis shows that the direct effects of digital transformation and technological innovation are more significant in less developed regions, while the substitution effect is stronger in economically developed regions. The findings suggest that policies promoting digital transformation and technological innovation should avoid a uniform approach and instead adopt coordinated and phased strategies that are suitable for regional development conditions. By providing new empirical evidence on the interaction between digital economy development and innovation investment, this study enriches the existing literature and offers policy implications for emerging economies seeking to achieve sustainable foreign trade development under increasing global trade uncertainty.

Accurate prediction of key environmental parameters is crucial for intelligent control and optimization, yet it remains challenging due to gradient instability in deep learning models, like Long Short-Term Memory (LSTM), during time series forecasting. This study introduces a novel adaptive optimization algorithm, NadamClip, which integrates gradient clipping directly into the Nadam framework to address the trade-off between convergence efficiency and gradient explosion. NadamClip incorporates an adjustable gradient clipping threshold strategy that permits manual tuning. Through systematic experiments, we identified an optimal threshold range that effectively balances model performance and training stability, dynamically adapting to the evolving convergence characteristics of the network across different training phases. Aquaculture systems are regarded as similar to modern biomanufacturing systems. The study evaluated an aquaculture dataset for ammonia concentration prediction in aquaculture environmental control processes. NadamClip achieved outstanding results on key metrics, including a Root Mean Square Error (RMSE) of 0.2644, a Mean Absolute Error (MAE) of 0.6595, and a Coefficient of Determination (R²) score of 0.9743. Compared to existing optimizer enhancements, NadamClip pioneers the integration of gradient clipping with adaptive momentum estimation, overcoming the traditional paradigm where clipping primarily serves as an external training control rather than an intrinsic algorithmic component. This study provides a practical and reproducible optimization framework for intelligent modeling of dynamic process systems, thereby contributing to the broader advancement of machine learning methods in predictive modeling and optimization for data-driven manufacturing and environmental processes.

Undoubtedly, time series data prediction stands as a primary focus of computational intelligence among researchers in both academia and industry, owing to its wide-ranging applications and significant impact. The efficacy of prediction models heavily relies on the optimization techniques employed to enhance both efficiency and accuracy. Within the realm of Machine Learning (ML), researchers have developed numerous optimization techniques and models, leading to a plethora of studies. Consequently, there exists a considerable body of literature comprising reviews of ML-based time series prediction. Recently, Deep Learning (DL) models have emerged in this domain, demonstrating performance levels that notably surpass those of traditional ML methods. Despite the burgeoning interest in advancing time series prediction models, there remains a noticeable absence of systematic review papers dedicated solely to enhancing the efficiency and accuracy of optimization techniques. Therefore, this paper is motivated by the need to present a systematic review of the efficiency and accuracy of optimizers studies concerning time series prediction implementations. We not only classify these studies based on their targeted optimization techniques, prediction applications—such as crop yield prediction, weather forecasting for farming, and pest detection and management—but also categorize them according to the types of optimization techniques and models employed, including Convolutional Neural Networks (CNNs), Deep Belief Networks (DBNs), and Long-Short Term Memory (LSTM) networks. Additionally, this study endeavor to provide insights into the future of the field by highlighting the challenges and potential future research opportunities, thereby offering guidance to interested researchers.

Accurate and dynamic prediction of water quality indicators is increasingly critical due to rising pollution and water resource insecurity, particularly when dealing with high-dimensional, nonlinear time series data. Dissolved oxygen (DO), a key indicator of aquatic ecosystem health and pollution, requires high prediction accuracy for effective environmental management. This study aims to enhance the accuracy and adaptability of DO prediction by addressing the limitations of traditional deep learning methods, such as slow convergence and local optima. We propose a novel hybrid optimization framework that combines Nesterov-accelerated Adaptive Moment Estimation (Nadam) with the differential evolution algorithm. A dual-population cooperation strategy and an information exchange mechanism were incorporated during the training of a long short-term memory (LSTM) network to achieve a dynamic balance between global exploration and local exploitation. This improves the model’s optimization efficiency and generalization. The research utilized a multivariate water quality time series dataset from Kaggle based on official data monitoring. Correlation analysis was conducted to ensure the scientific validity and effectiveness of the selected input variables. Experimental results demonstrated that the proposed method significantly outperforms traditional optimization strategies for DO prediction. Compared to the original Nadam optimizer, it reduced the mean squared prediction error by 47.8%, exhibiting enhanced adaptability and robustness in complex pollution scenarios. This study presents an effective optimization strategy to improve LSTM performance in water quality forecasting, along with a scalable and interpretable intelligent analysis framework. It provides both theoretical and practical support for water quality forecasting, early warning systems, and intelligent environmental monitoring.