L. Sujihelen
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
L. Sujihelen has a total of 39 co-authors. Here are the top co-authors by number of shared publications:
C. Senthilsingh
Shadan Womens College of Engineering and Technology
3 shared publicationsC. Jayakumar
Sri Venkateswara College of Engineering
2 shared publicationsC. Senthil Singh
Department of Computer Science and Engineering, hindusthan institute of technology Valley Campus, Coimbatore, India
2 shared publications- Manojkumar Kothandapani
Department of Computer Science and Engineering, Sathyabama Institute of Science and Technology, Chennai, India
2 shared publications A. Sivasangari
Sathyabama Institute of Science and Technology
1 shared publication- Anandakumar Haldorai
Sri Eshwar College of Engineering
1 shared publication
Latest Publications
- Open Access
- 20 November 2026
- Published by Springer Nature in Discover Computing
- Vol. 29 (1)
Privacy-preserving federated digital twin-assisted edge healthcare monitoring
- 20 November 2026
- Published by Springer Nature in Discover Computing
- Vol. 29 (1)
Open AccessPrivacy-preserving federated digital twin-assisted edge healthcare monitoring
- 11 September 2026
- Published by World Scientific Pub Co Pte Ltd in International Journal of Image and Graphics
C2H-ADLTM: Adaptive Distributed Multi-Objective Deep Learning Model for Missing Data Imputation in Healthcare
Missing data refers to the absence of data points within the dataset, which can occur for several reasons and creates diverse impacts in healthcare applications. One of the frequently used methods for handling missing values is imputation techniques, which estimate and substitute missing values with probable ones based on known data points. However, these techniques may not incorporate complex medical data relationships, create training difficulties, and lack generalization ability. Consequently, to overcome these shortcomings, this research initiated a Crocuta Community Hunt Optimization-enabled adaptive distributed learning-based Bi-directional Long Short-Term Memory (C2H-ADLTM). The adaptive distributed learning concept enables the C2H-ADLTM model to attain faster learning performance and handle long-term dependencies. The designed C2H algorithm is deployed to fine-tune the hyperparameters of the proposed model for handling large and complex datasets with minimal training time. Moreover, the research utilized a multi-objective function, which emphasizes the robustness of the C2H algorithm for dealing with several types of missing data and ensures efficient as well as accurate imputation. When compared with the progressive methods, the experimental outcomes show that the proposed model attains superior imputation performance with a reduced MAE of 4.69, MSE of 4.83, Median absolute error of 2.76, [Formula: see text] score of 0.87, and RMSE of 4.34 for the Pima Indian diabetes dataset.
- 11 September 2026
- Published by World Scientific Pub Co Pte Ltd in International Journal of Image and Graphics
C2H-ADLTM: Adaptive Distributed Multi-Objective Deep Learning Model for Missing Data Imputation in Healthcare
Missing data refers to the absence of data points within the dataset, which can occur for several reasons and creates diverse impacts in healthcare applications. One of the frequently used methods for handling missing values is imputation techniques, which estimate and substitute missing values with probable ones based on known data points. However, these techniques may not incorporate complex medical data relationships, create training difficulties, and lack generalization ability. Consequently, to overcome these shortcomings, this research initiated a Crocuta Community Hunt Optimization-enabled adaptive distributed learning-based Bi-directional Long Short-Term Memory (C2H-ADLTM). The adaptive distributed learning concept enables the C2H-ADLTM model to attain faster learning performance and handle long-term dependencies. The designed C2H algorithm is deployed to fine-tune the hyperparameters of the proposed model for handling large and complex datasets with minimal training time. Moreover, the research utilized a multi-objective function, which emphasizes the robustness of the C2H algorithm for dealing with several types of missing data and ensures efficient as well as accurate imputation. When compared with the progressive methods, the experimental outcomes show that the proposed model attains superior imputation performance with a reduced MAE of 4.69, MSE of 4.83, Median absolute error of 2.76, [Formula: see text] score of 0.87, and RMSE of 4.34 for the Pima Indian diabetes dataset.
- Open Access
- 18 August 2025
- Published by Springer Nature in International Journal of Computational Intelligence Systems
- Vol. 18 (1) , 1-20
AQST-ClustNet: Hybrid Aquila Quantum Sooty Tern Optimization for User Profile Clustering in Social Network
User profile clustering is the process of grouping users on social media sites based on common characteristics identified in their profile data, such as demographics, interests, and interactions. Profile clustering allows users to engage in targeted marketing, skill-matching, and collaborative networking by grouping them based on similar attributes, interests, or professional criteria. However, one key drawback of user profile clustering is its sensitivity to noisy and missing data, high-dimensional feature spaces, poor semantic understanding, and complexity limitations. To overcome these issues, a novel Hybrid Aquila Quantum Sooty tern optimization for clustering (AQST-ClustNet) approach based on user profiles (UP) has been proposed in this paper. The user profile data is preprocessed using NLP techniques involving data stemming, handling of missing data or values, removal of stop words, and data extraction for eliminating inappropriate data. A Hybrid Aquila Quantum Sooty Tern Optimization (HAQSTO) algorithm is employed for clustering the user profile into healthcare professionals, marketing professionals, software developers, and educators. The efficiency of the developed method is assessed employing various metrics, including Calinski–Harabasz score (CHS), Silhouette score (SHS), and Davies–Bouldin score (DBS). The proposed model achieves less runtime of 45 s, whereas the existing techniques, such as MCEMS, DBSTexC, and TSMIUSC-Miner, achieve runtimes of 70 s, 79 s, and 60 s. Using the effective dual-stage feature extraction and clustering approach, the complexity of clustering and a high-dimensional feature space is effectively reduced.
- 18 August 2025
- Published by Springer Nature in International Journal of Computational Intelligence Systems
- Vol. 18 (1) , 1-20
Open AccessAQST-ClustNet: Hybrid Aquila Quantum Sooty Tern Optimization for User Profile Clustering in Social Network
User profile clustering is the process of grouping users on social media sites based on common characteristics identified in their profile data, such as demographics, interests, and interactions. Profile clustering allows users to engage in targeted marketing, skill-matching, and collaborative networking by grouping them based on similar attributes, interests, or professional criteria. However, one key drawback of user profile clustering is its sensitivity to noisy and missing data, high-dimensional feature spaces, poor semantic understanding, and complexity limitations. To overcome these issues, a novel Hybrid Aquila Quantum Sooty tern optimization for clustering (AQST-ClustNet) approach based on user profiles (UP) has been proposed in this paper. The user profile data is preprocessed using NLP techniques involving data stemming, handling of missing data or values, removal of stop words, and data extraction for eliminating inappropriate data. A Hybrid Aquila Quantum Sooty Tern Optimization (HAQSTO) algorithm is employed for clustering the user profile into healthcare professionals, marketing professionals, software developers, and educators. The efficiency of the developed method is assessed employing various metrics, including Calinski–Harabasz score (CHS), Silhouette score (SHS), and Davies–Bouldin score (DBS). The proposed model achieves less runtime of 45 s, whereas the existing techniques, such as MCEMS, DBSTexC, and TSMIUSC-Miner, achieve runtimes of 70 s, 79 s, and 60 s. Using the effective dual-stage feature extraction and clustering approach, the complexity of clustering and a high-dimensional feature space is effectively reduced.
- 9 May 2025
- Published by Springer Nature in Signal, Image and Video Processing
- Vol. 19 (7) , 1-12
Monitoring the crowd emotion using valence and arousal of crowd based on prominent features of crowd
- 9 May 2025
- Published by Springer Nature in Signal, Image and Video Processing
- Vol. 19 (7) , 1-12
Monitoring the crowd emotion using valence and arousal of crowd based on prominent features of crowd
- Open Access
- 11 February 2025
- Published by Poltekkes Kemenkes Surabaya in Journal of Electronics, Electromedical Engineering, and Medical Informatics
- Vol. 7 (2) , 352-365
Categorizing Crowd Emotions based on Cross Division Expressions and Anomalies
The crowd emotion sensing is a critical element in surveillance and management of the crowd in different environments. With exploding populations, and developing nations, the crowd in urban cities mandate state of art surveillance methodologies involving continuous monitoring and reporting of criminal activities. The research article presents a novel technique to compute the spatial and temporal features obtained from the crowd environments and combine the novelty of neural networks for detecting the emotions of crowds with better accuracy and swiftness. The features are obtained from the continuous feed of surveillance videos typically categorized into the common features of human beings namely anger, sadness, disgust, surprise, fear, happiness and obviously neutrality. Such features are extracted after careful background separation which are typically difficult in crowded environments, using techniques namely SIFT, and FAST termed to be the visual descriptors. Once the features are extracted, spatial and temporal features are classified into individual and combined features as defined in the cross-division environment in order to portray the crowd dynamics and characteristics. Cross division environment computes the necessary features for identifying the anomalies in the crowded situations in a neural network, after a series of operations such as dimensionality reduction, and principal component analysis. From the semantic information, crowd behaviours are detected based on interactive features in a dynamic environment and the proposed technique has demonstrated effective results in terms of 98.9% accuracy in detecting especially violence in crowd datasets collected from UMN.
- 11 February 2025
- Published by Poltekkes Kemenkes Surabaya in Journal of Electronics, Electromedical Engineering, and Medical Informatics
- Vol. 7 (2) , 352-365
Open AccessCategorizing Crowd Emotions based on Cross Division Expressions and Anomalies
The crowd emotion sensing is a critical element in surveillance and management of the crowd in different environments. With exploding populations, and developing nations, the crowd in urban cities mandate state of art surveillance methodologies involving continuous monitoring and reporting of criminal activities. The research article presents a novel technique to compute the spatial and temporal features obtained from the crowd environments and combine the novelty of neural networks for detecting the emotions of crowds with better accuracy and swiftness. The features are obtained from the continuous feed of surveillance videos typically categorized into the common features of human beings namely anger, sadness, disgust, surprise, fear, happiness and obviously neutrality. Such features are extracted after careful background separation which are typically difficult in crowded environments, using techniques namely SIFT, and FAST termed to be the visual descriptors. Once the features are extracted, spatial and temporal features are classified into individual and combined features as defined in the cross-division environment in order to portray the crowd dynamics and characteristics. Cross division environment computes the necessary features for identifying the anomalies in the crowded situations in a neural network, after a series of operations such as dimensionality reduction, and principal component analysis. From the semantic information, crowd behaviours are detected based on interactive features in a dynamic environment and the proposed technique has demonstrated effective results in terms of 98.9% accuracy in detecting especially violence in crowd datasets collected from UMN.
- 29 September 2024
- Published by Springer Nature
- p. 428-441
Real-Time Phonic Decipherer
- Retracted
- 15 June 2023
- Published by Springer Nature in Soft Computing
- p. 1-15
RETRACTED ARTICLE: Type 2 diabetes mellitus classification using predictive supervised learning model
- 15 June 2023
- Published by Springer Nature in Soft Computing
- p. 1-15
RetractedRETRACTED ARTICLE: Type 2 diabetes mellitus classification using predictive supervised learning model
- 4 June 2022
- Published by Springer Nature
- p. 639-648
Smart Agriculture Framework Implemented Using the Internet of Things and Deep Learning
- 4 June 2022
- Published by Springer Nature
- p. 639-648
Smart Agriculture Framework Implemented Using the Internet of Things and Deep Learning
- Open Access
- 1 January 2022
- Published by Wiley in Wireless Communications and Mobile Computing
- Vol. 2022 (1)
Node Replication Attack Detection in Distributed Wireless Sensor Networks
Wireless sensor network (WSN) is an emerging technology used in emergency scenarios. There are a number of possible threats to WSNs because they use unsupervised IP addresses. Securing networks with unattended sensors is a real challenge nowadays. Sensor nodes lack power and storage, making them incompatible with normal security checks. It will be vital to make advancements in sensor network architecture and protocol design. There will be more vulnerability to attack if there is a lack of security. Especially, one key attack is node replication which induces the sensor node to acts as an original node, collecting data from the network and sending it to the attacker. In dynamic WSN, detecting an assault is difficult to find replica nodes. Therefore, this paper proposes a Strategic Security System (SSS) to discover replica nodes in static and dynamic distributed WSNs. It is mainly focused on enhancing detection accuracy, time delay, and communication overhead. The present system includes Single Stage Memory Random Walk with Network Division (SSRWND) and a Random‐walk‐based approach to detect clone attacks (RAWL). The proposed system has less memory and better detection accuracy.
- 1 January 2022
- Published by Wiley in Wireless Communications and Mobile Computing
- Vol. 2022 (1)
Open AccessNode Replication Attack Detection in Distributed Wireless Sensor Networks
Wireless sensor network (WSN) is an emerging technology used in emergency scenarios. There are a number of possible threats to WSNs because they use unsupervised IP addresses. Securing networks with unattended sensors is a real challenge nowadays. Sensor nodes lack power and storage, making them incompatible with normal security checks. It will be vital to make advancements in sensor network architecture and protocol design. There will be more vulnerability to attack if there is a lack of security. Especially, one key attack is node replication which induces the sensor node to acts as an original node, collecting data from the network and sending it to the attacker. In dynamic WSN, detecting an assault is difficult to find replica nodes. Therefore, this paper proposes a Strategic Security System (SSS) to discover replica nodes in static and dynamic distributed WSNs. It is mainly focused on enhancing detection accuracy, time delay, and communication overhead. The present system includes Single Stage Memory Random Walk with Network Division (SSRWND) and a Random‐walk‐based approach to detect clone attacks (RAWL). The proposed system has less memory and better detection accuracy.
- Open Access
- 28 April 2021
- Published by Springer Nature in Research Square
Detecting Replica Node in Distributed Mobile Wireless Sensor Networks
Security is an important problem in wireless sensor networks. There are more attacks in WSN. Node replication attack is an important attack in wireless sensor networks which can be easily captured by adversaries. In this attack, the node will behave as an original node and collect all the information which is transferred in the network and passes to the attacker node. Some existing schemes are proposed for predicting the replica nodes. In existing method, the detection rate is less, communication cost is high. To increase the detection accuracy and to increase the communication cost, a History of Neighbor Node (HNN) method is proposed. The proposed HNN approach detects the replicated node locally and globally. The HNN method detection accuracy is high in any speed limit. Another approach also proposed for static mobile sensor networks, FEC approach, the replicated node is detected based upon the speed, direction and location. In random time the base station verifies the node with the observed speed and direction. The verification is conducted by the SPRT technique. The SPRT technique tests the node which should be rejected (clone node) or accepted (genuine node). Once the clone node is found quickly, then it is removed from the network. The proposed system has high detection accuracy and less communication cost and less energy efficiency when compared with the existing system.
- 28 April 2021
- Published by Springer Nature in Research Square
Open AccessDetecting Replica Node in Distributed Mobile Wireless Sensor Networks
Security is an important problem in wireless sensor networks. There are more attacks in WSN. Node replication attack is an important attack in wireless sensor networks which can be easily captured by adversaries. In this attack, the node will behave as an original node and collect all the information which is transferred in the network and passes to the attacker node. Some existing schemes are proposed for predicting the replica nodes. In existing method, the detection rate is less, communication cost is high. To increase the detection accuracy and to increase the communication cost, a History of Neighbor Node (HNN) method is proposed. The proposed HNN approach detects the replicated node locally and globally. The HNN method detection accuracy is high in any speed limit. Another approach also proposed for static mobile sensor networks, FEC approach, the replicated node is detected based upon the speed, direction and location. In random time the base station verifies the node with the observed speed and direction. The verification is conducted by the SPRT technique. The SPRT technique tests the node which should be rejected (clone node) or accepted (genuine node). Once the clone node is found quickly, then it is removed from the network. The proposed system has high detection accuracy and less communication cost and less energy efficiency when compared with the existing system.