Ahmed et al. (2025) Efficient Hybrid Anomaly Detection in Environmental Data

Identification

• Journal: Lecture notes in networks and systems
• Year: 2025
• Date: 2025-11-20
• Authors: Muhammad R. Ahmed, Mohammed Aseeri, M. Y. O. Thirein, M. Shamim Kaiser, Ifat Al Baqee, Mohammad Hamiruce Marhaban
• DOI: 10.1007/978-981-95-1069-6_11

Research Groups

• Military Technological College, Muscat, Oman
• King Abdulaziz City for Science and Technology (KACST), Riyadh, Saudi Arabia
• Institute of Information Technology, Jahangirnagar University, Savar, Bangladesh
• School of Science and Engineering, Canadian University of Bangladesh, Dhaka, Bangladesh
• Faculty of Engineering, Universiti Putra Malaysia, Serdang, Selangor, Malaysia

Short Summary

This study develops an efficient hybrid anomaly detection model by combining K-Nearest Neighbors (KNN) and Isolation Forests (IF) to identify unusual patterns in environmental monitoring system data. Applied to a dataset of 56,996 points, the model identified 2,872 anomalous patterns with an approximate detection accuracy of 5.44%.

Objective

• To develop and apply an efficient hybrid anomaly detection algorithm, combining K-Nearest Neighbors (KNN) and Isolation Forests (IF), for identifying unusual patterns in environmental monitoring data that may indicate environmental changes or sensor malfunctions.

Study Configuration

• Spatial Scale: Environmental monitoring data from Lake Macquarie City Council, Australia.
• Temporal Scale: Data collected historically up to 31 July 2020.

Methodology and Data

• Models used: K-Nearest Neighbors (KNN), Isolation Forests (IF), and a hybrid algorithm combining both.
• Data sources: Environmental monitoring system (EMS) sensors developed by ARCS Group in partnership with the University of Technology, Sydney, as part of the TULIP Project. Data was collected by the Lake Macquarie City Council using their Community IoT LoRaWAN Network—The Things Network.

Main Results

• The hybrid model was applied to a dataset comprising 56,996 data points.
• It identified 2,872 anomalous patterns from a subset of 52,978 data points.
• The detection accuracy achieved was approximately 5.44%.
• The combination of KNN and IF enhanced the detection accuracy and robustness of the hybrid model.

Contributions

• Proposes a novel hybrid anomaly detection approach that integrates K-Nearest Neighbors and Isolation Forests for environmental data.
• Demonstrates improved detection accuracy and robustness compared to individual methods (implied by augmentation).
• Offers valuable insights into the characteristics and distribution of anomalous patterns in environmental monitoring.
• Addresses practical limitations in environmental monitoring, such as computational complexity and the need for heuristic threshold setting.
• Provides a reliable method for anomaly detection, thereby contributing to the enhancement of environmental data analysis and monitoring.

Funding

• The data used in this study originated from environmental monitoring system (EMS) sensors developed as part of the TULIP Project. No direct funding for the research presented in the paper is explicitly mentioned.

Citation

@article{Ahmed2025Efficient,
  author = {Ahmed, Muhammad R. and Aseeri, Mohammed and Thirein, M. Y. O. and Kaiser, M. Shamim and Baqee, Ifat Al and Marhaban, Mohammad Hamiruce},
  title = {Efficient Hybrid Anomaly Detection in Environmental Data},
  journal = {Lecture notes in networks and systems},
  year = {2025},
  doi = {10.1007/978-981-95-1069-6_11},
  url = {https://doi.org/10.1007/978-981-95-1069-6_11}
}