All reviews of published articles are made public. This includes manuscript files, peer review comments, author rebuttals and revised materials. Note: This was optional for articles submitted before 13 February 2023.
Peer reviewers are encouraged (but not required) to provide their names to the authors when submitting their peer review. If they agree to provide their name, then their personal profile page will reflect a public acknowledgment that they performed a review (even if the article is rejected). If the article is accepted, then reviewers who provided their name will be associated with the article itself.
Dear authors, thank you for fixing the minor issues reported in the previous revision round.
[# PeerJ Staff Note - this decision was reviewed and approved by Massimiliano Fasi, a PeerJ Section Editor covering this Section #]
The revised manuscript reflects significant improvements in both clarity and technical depth. Specific enhancements include a more detailed explanation of the HABCOeSNN architecture, the inclusion of computational cost considerations, and a clearer breakdown of the contributions of the ABC algorithm and the evolving spiking neural network components. Additionally, the manuscript now includes valuable discussions on scalability, potential real-time deployment challenges, and future work directions such as adaptation to concept drift and domain extension.
The authors have demonstrated a strong experimental design that effectively validates the proposed HABCOeSNN model. By employing two diverse benchmark datasets—Numenta Anomaly Benchmark (NAB) and Yahoo Webscope—and comparing the results against five leading metaheuristic optimization algorithms (PSO, FPA, GWO, WOA, and GS) as well as traditional classifiers (Random Forest, SVM, and k-NN), the study offers a comprehensive evaluation framework.
The findings presented in the manuscript appear to be valid and well-supported by thorough experimentation and analysis. The authors have demonstrated a strong commitment to ensuring the reliability of their results through a carefully constructed evaluation strategy.
Paper can be accepted.
no comment
no comment
no comment
no comment
Dear Authors,
The reviewers find the paper to be a valuable contribution to the field of unsupervised anomaly detection in streaming data. As reviewers reported, the model demonstrates superior performance on benchmark datasets and is supported by rigorous statistical analysis.
However, minor revisions are still necessary. Specifically, the literature review should be updated with more recent work and organized thematically. Also, the abstract should include quantitative results and provide a more precise explanation of the approach's novelty.
More detail is needed on experimental settings, including hardware and software configurations, to ensure reproducibility and clarity on the fairness of comparisons with other methods.
Reviewers also recommend improving abbreviation consistency, elaborating on conclusions, and enhancing the explanation of practical applicability and scalability. Additional insights into the individual contributions of the hybrid components and an expanded evaluation across diverse datasets are also encouraged.
**PeerJ Staff Note:** Please ensure that all review, editorial, and staff comments are addressed in a response letter and that any edits or clarifications mentioned in the letter are also inserted into the revised manuscript where appropriate.
**Language Note:** The review process has identified that the English language must be improved. PeerJ can provide language editing services - please contact us at [email protected] for pricing (be sure to provide your manuscript number and title). Alternatively, you should make your own arrangements to improve the language quality and provide details in your response letter. – PeerJ Staff
1. The paper addresses the challenge of unsupervised anomaly detection in streaming data, which is crucial for real-time applications in smart devices. It highlights the issue of manual hyperparameter tuning in evolving spiking neural networks (eSNNs).
2. The authors introduce a novel hybrid approach called HABCOeSNN, which integrates the Artificial Bee Colony (ABC) algorithm with Online Evolving Spiking Neural Networks (OeSNN) to enhance performance by optimizing hyperparameters.
3. The model is tested on two benchmark datasets—Numenta Anomaly Benchmark (NAB) and Yahoo Webscope—and compared against five well-established optimization algorithms (PSO, FPA, GWO, WOA, and GS) as well as traditional classifiers like Random Forest, SVM, and k-NN.
4. The HABCOeSNN model outperforms other methods in accuracy and anomaly detection, as demonstrated through F1-score, precision, recall, and balanced accuracy. The study also confirms statistical significance using a one-way ANOVA test.
5. The study sets a new standard in unsupervised anomaly detection, proving the importance of metaheuristic-based hyperparameter optimization for evolving neural networks in handling dynamic streaming data.
1. Two diverse benchmark datasets (NAB and Yahoo Webscope) were chosen, which include real-world and synthetic time-series data with imbalanced anomaly distributions.
2. The proposed model is compared against five optimization algorithms (PSO, FPA, GWO, WOA, and GS) and three machine learning classifiers (Random Forest, SVM, and k-NN), ensuring a broad assessment.
3. Three different configurations (Set 1, Set 2, Set 3) are used to evaluate the impact of hyperparameter tuning on model performance.
4. Performance is assessed using multiple metrics, including F1-score, balanced accuracy (BA), precision, and recall, making the findings more reliable.
5. A one-way ANOVA test is conducted to confirm that the observed performance differences are statistically significant (p < 0.05), strengthening the credibility of the results.
1. The use of both real-world and synthetic datasets enhances the generalizability of the model's findings across various anomaly detection scenarios.
2. The study evaluates the proposed method against 14 state-of-the-art anomaly detection algorithms, ensuring that the observed improvements are not coincidental.
3. The results show that Set 2 consistently outperforms other configurations, providing strong evidence that optimized hyperparameters enhance model performance.
4. The study uses multiple evaluation measures (F1-score, BA, precision, recall) and heatmaps to visually confirm model robustness.
5. The one-way ANOVA test confirms that the performance improvements of HABCOeSNN are statistically significant, adding reliability to the conclusions drawn.
• The study focuses on streaming data but does not explicitly discuss the scalability of HABCOeSNN in high-volume, real-time processing scenarios.
• While the model improves accuracy, the computational cost of combining ABC with eSNN should be further analyzed.
• The study lacks discussion on practical deployment challenges, such as integration with IoT devices or cloud-based anomaly detection systems.
• More datasets from different domains, such as cybersecurity or financial fraud detection, could further validate the model’s robustness.
• Since streaming data evolves over time, a deeper analysis of how HABCOeSNN adapts to long-term concept drift would be beneficial.
• A breakdown of the individual contributions of ABC optimization and eSNN evolution would help clarify which component contributes the most to performance gains.
1) Current articles in the literature should also be used! The literature section should be updated.
2)The abstract section should be supported by numerical information.
1) What is the novelty of this study? How does it contribute to the literature? It should be explained.
1) Here, the proposed method is compared with different metaheuristic algorithms. It should be stated which control parameters are used for both the proposed method and other metaheuristic algorithms? Is there a fair comparison environment?
1) The conclusion section does not adequately reflect the study. It should be expanded to provide the general framework.
This paper proposes HABCOeSNN, a hybrid anomaly detection method that integrates the Artificial Bee Colony (ABC) algorithm with an Online Evolving Spiking Neural Network (OeSNN), aiming to improve real-time unsupervised anomaly detection in streaming data. The method is evaluated on standard benchmarks and shows superior performance over multiple optimization strategies. The suggestions can be considered:
1. The abstract briefly mentions the proposed HABCOeSNN framework, but lacks detail on how it operates or why it is effective.
2. Throughout the manuscript, please ensure consistent use of full terms and their abbreviations. For example, “Spiking Neural Network (SNN)”, “Evolving Spiking Neural Network (eSNN)”, and “Online Evolving Spiking Neural Network (OeSNN)” should be clearly defined once and used consistently thereafter. Perform a thorough check for abbreviation consistency in the full text.
3. The related work section presents many methods but lacks synthesis. Instead of listing papers individually, group and summarize them by category.
4. The overall presentation would benefit from a careful proofreading pass focused on the formatting of formulas and mathematical symbols.
5. To further substantiate the significance of the proposed method, it is recommended to include advanced anomaly detection techniques from different fields, such as: frequency-to-spectrum mapping GAN; non-local and local feature-coupled self-supervised network; learning prompt-enhanced context features; anomaly detection for medical images using heterogeneous auto-encoder.
The experimental section lacks basic implementation details. Please specify the hardware and software environments used for training and evaluation. This information is necessary for reproducibility.
no comment
no comment
All text and materials provided via this peer-review history page are made available under a Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.