Application of Improved Particle Swarm Optimization SVM in Water Quality Evaluation of Ming Cui Lake

Zunyang Zhang, Cheng Yang, Qiao Qiao, Xuesheng Li, Fuping Wang, Chengcheng LiSustainability 2023 · DOI

TL;DR

This paper introduces a novel CPOS-SVM model for water quality evaluation that significantly reduces training time by using Pareto optimality for data sparsification and optimizes SVM parameters with a Particle Swarm Optimization (PSO) algorithm. The proposed method achieves high accuracy (94%) and remarkable speed, making it a valuable tool for environmental monitoring and management.

What Problem Does This Paper Address?

Accurate water quality assessment is crucial for environmental protection and resource management. Traditional methods often require extensive data, leading to long computation times for evaluation models. Furthermore, Support Vector Machines (SVMs), a popular choice for such tasks, are sensitive to parameter selection (kernel parameter 'g' and penalty factor 'c'), making it challenging to achieve optimal accuracy. This paper tackles these issues by proposing an improved SVM model that addresses both computational efficiency and parameter optimization for water quality evaluation.

Methods

The core of this research lies in the CPOS-SVM (Classification with Pareto Optimal Solution-SVM) model. It combines two key innovations:

  • Data Sparsification using Pareto Optimality: To combat the computational burden of large datasets, the authors employ the concept of Pareto optimality. This technique sparsifies the training data by identifying and retaining only the most representative "support vectors" that define the decision boundaries. This process aims to reduce the training set size without losing critical data characteristics, thereby speeding up training.
  • Parameter Optimization with Particle Swarm Optimization (PSO): The performance of SVMs is heavily dependent on the kernel parameter 'g' and the penalty factor 'c'. The paper utilizes a PSO algorithm to efficiently search for the optimal combination of these hyperparameters. PSO is an intelligent optimization algorithm known for its ability to find approximate optimal solutions with high accuracy and simple implementation, making it suitable for tuning SVM parameters.

The overall workflow involves:

  1. Data Preprocessing: Including Pareto optimal solution-based sparsification and normalization.
  2. PSO-based Parameter Optimization: Finding optimal 'g' and 'c' values for the SVM.
  3. SVM Model Training: Using the sparsified training data and optimized parameters.
  4. Model Evaluation: Testing the trained model on unseen data.

Datasets

The study utilizes water quality monitoring data collected from the Ming Cui Wetland Park, Yinchuan, Ningxia, China, spanning from 2014 to 2021. To augment the dataset and test the model's scalability, the authors generated synthetic datasets by interpolating between adjacent water quality standards. This resulted in four distinct dataset sizes: 620, 1020, 1520, and 2020 groups of sample data. These datasets were randomly split into training and testing sets for model development and evaluation.

Key Insights and Ideas

  • Pareto Optimality as a Data Compressor: The application of Pareto optimality for data sparsification is a clever way to reduce the computational load of SVMs without sacrificing essential data features. This is particularly relevant for environmental monitoring where data can accumulate rapidly.
  • PSO as an Efficient SVM Tuner: Leveraging PSO to optimize SVM hyperparameters (g and c) offers a more systematic and efficient approach compared to traditional grid search or manual tuning, especially when dealing with complex parameter spaces.
  • Speed-Accuracy Trade-off Management: The CPOS-SVM model demonstrates that it's possible to achieve both significant speed improvements and maintain high accuracy, challenging the notion that faster models must compromise on performance.
  • Robustness with Increasing Data: The study suggests that while the initial advantage of data sparsification might be less pronounced with smaller datasets, its benefits become more apparent as the data volume grows, leading to more stable and robust performance.

Key Results

The CPOS-SVM model demonstrated impressive performance:

  • Training Time: The model achieved a training time of under 2 seconds, which remained constant even with increasing data volume. This is a significant improvement over traditional SVM and other comparative models.
  • Accuracy: The CPOS-SVM model achieved a prediction accuracy of approximately 94%. This was comparable to the POS-SVM model and slightly better than the standard SVM and POS-BP models.
  • Comparative Performance:
    • SVM: Fastest training time when parameters are pre-set, but training time increases dramatically with data volume.
  • POS-SVM: High accuracy, but training time increases with data volume.
  • POS-BP: Accuracy increased significantly with data volume, but generally lower than CPOS-SVM.
  • CPOS-SVM: Achieved the shortest training time and maintained high accuracy, showing strong stability and robustness.

Limitations and Future Work

  • Generalizability: While the study focused on Ming Cui Lake, further validation on diverse water bodies with varying pollution characteristics would be beneficial to confirm the model's broader applicability.
  • Parameter Sensitivity: Although PSO optimizes parameters, the inherent sensitivity of SVM to hyperparameter choices might still be a factor in specific, highly complex scenarios.
  • Future Work: The authors suggest that this model has high application value in water quality evaluation and can contribute to better environmental protection and resource management. Future work could explore integrating this model into real-time monitoring systems or extending it to predict other environmental parameters.

Summary updated

Abstract

Water quality directly determines our living environment. In order to establish a more scientific and reasonable water quality evaluation model, it needs a lot of data support, but it will lead to a large increase in the calculation time of the evaluation model. This paper proposes an improved particle swarm optimization SVM model (CPOS-SVM) to solve this problem. In this paper, the Pareto optimal solution concept is used to sparsely process the training set, which can ensure that the number of training sets is reduced without loss of data characteristics, thus reducing the training time. In order to solve the problem of the kernel parameter g and penalty factor c on the SVM algorithm, which affects the accuracy of the SVM model but it is difficult to select why, a particle swarm optimization algorithm is used in this paper to optimize the kernel parameter and penalty factor and improve the accuracy of the model. In this paper, 480 sets of data from Ming Cui Lake from 2014 to 2022 are taken as the research object, and examples are analyzed in MATLAB 2020a. The results show that the training time of the CPOS-SVM model can be completed within 2 s and does not increase with the increase of data volume. Meanwhile, by comparing the SVM model, POS-SVM model, and POS-BP model, training time increases dramatically with the amount of data. The accuracy of the POS-SVM model is the highest, and the accuracy of the CPOS-SVM model is basically consistent with that of the POS-SVM, reaching 94%, while the accuracy of the SVM model and the POS-BP model are slightly worse. This indicates that the CPOS-SVM model has good application value in water quality evaluation.
BibTeX
@inproceedings{zhang2023application,
  author = {Zunyang Zhang and Cheng Yang and Qiao Qiao and Xuesheng Li and Fuping Wang and Chengcheng Li},
  title = {{Application of Improved Particle Swarm Optimization SVM in Water Quality Evaluation of Ming Cui Lake}},
  booktitle = {Sustainability},
  year = {2023},
  doi = {10.3390/su15129835},
}
Figure