Institute of Public Enterprise

The proposed control strategy synergistically integrates Model Reference Adaptive Control (MRAC) with Field-Oriented Control (FOC) to achieve highly precise and dynamic torque and magnetic flux in induction motor propulsion units. It effectively compensates for nonlinearities and parameter variations under dynamic operating conditions. Integration with a seven-level CHB multilevel inverter employing SiC MOSFETs enhances efficiency, switching speed, and voltage quality. Unlike conventional FOC, which is sensitive to parameter drift due to thermal and magnetic effects, the adaptive law, derived through Lyapunov stability theory, ensures robust speed tracking despite disturbances. Simulation results confirm improved transient performance, reduced torque ripple, superior tracking accuracy, and lower total harmonic distortion (THD). This architecture offers a reliable, high-efficiency solution for advanced electric vehicle (EV) propulsion systems.

Breast cancer, a common disease in women globally, is caused by abnormal cell growth in the breast, forming tumors with potential fatality if not detected early. Breast cancer, a major global cause of death, necessitates the development of early-stage diagnostic tests. In this article, the authors show the Enhanced Breast Cancer Histopathology Vectorized Detection (EBCHVD) model that combines deep learning with topological data analysis in order to differentiate breast tissue samples as benign or malignant. For its detailed images, the model takes the Break His 400× histopathology dataset and applies preprocessing techniques followed by a meta-learning ensemble made up of different Convolutional Neural Networks (CNNs) to capture diverse image representations. the performance analysis is benign-malignant of f24 ≈ 0.45, f30 ≈ 0.12, loss of 0.70, accuracy of 0.65, confusion matrix of 86.5%, and ROC curve of 81% calculations.

Working with high-dimensional data is challenging, and the infrequency of fraudulent transactions complicates the detection of credit card fraud, posing a significant issue in the banking sector. This study introduces a thorough and lucid ensemble-based approach that integrates resampling techniques, feature selection, and interpretability to efficiently detect fraud in extremely imbalanced datasets. This study utilized the Kaggle Credit Card Fraud dataset to evaluate the efficacy of our effort. A total of 284,807 transactions occurred, with merely 0.17 % deemed fraudulent. To address the issue of uneven class distribution, this study employed SMOTE, ADASYN, and SMOTE-ENN. This study employed correlation analysis and mutual information selection to identify the most beneficial features. This study trained and evaluated three models: XGBoost, a Stacking Ensemble, and Random Forest. This study employed accuracy, precision, recall, F1-score, ROCAUC, and confusion matrices for this analysis. The Stacking Ensemble performed exceptionally on all assessments, indicating its efficacy in detecting elusive fraudulent transactions. SHAP-based explainability elucidated the significance of each feature on both global and local scales, hence enhancing the model's comprehensibility. The proposed architecture offers a transparent, precise, and dependable method for detecting credit card fraud within actual financial systems.

Purpose The purpose of this paper is to explore the role and effect of disclaimer dimensions in digital communication specifically in enhancing user engagement, perceived credibility and fostering innovation in digital communication. The study is motivated by the ongoing debate on whether disclaimers operate primarily through central or peripheral processing routes in the Elaboration Likelihood Model (ELM) framework and how these effects interact with digital literacy. Design/methodology/approach Survey was conducted to collect data from 2,980 participants within eight major cities, who have recently encountered a digital disclaimer. The results were obtained through a covariance-based structural equation modelling (CB-SEM) with SPSS AMOS and moderation analysis with the PROCESS macro. Findings The analytical results reveal that if disclaimers are clearly visible, comprehensive and culturally sensitive they lead to better user engagement and trustworthiness. Moderation analysis reveals the role of digital literacy and concludes that higher digital literacy helps online content users to understand the disclaimer. Originality/value This study is one of its own kind that focuses on the role of disclaimers in leading towards innovation in digital communication. This research is helpful for the policy makers to develop digital literacy initiatives and the design of effective disclaimers as it empirically tested both central and peripheral processing routes of disclaimers in a high-usage digital content environment.

As the number of vehicles on roads, expressways, and highways grows rapidly, efficient vehicle detection, classification, tracking and counting have become essential to the civilian, military, and governmental uses like highway monitoring, toll collection, traffic planning and flow regulation. Vehicle detection is one of the most important of these tasks, which can help in enabling effective traffic management. A solution to this is a Computer Vision (CV)based approach, which can be implemented without interfering with traffic and allows the adaption of the system. The paper is an introduction of a cost-efficient and mobile system based on CV methods to identify and count vehicles in real-time. The system works with the video frames to detect the moving vehicles with the help of background subtraction, which is followed by additional processing that groups the vehicles into the groups of light vehicles, heavy vehicles, and motorcycles.

CVDs are the major cause of death around the world, and intelligent diagnostics systems to identify the risks early are required. The traditional diagnostic procedures do not frequently work with complex and high-dimensional medical data. In order to address this drawback, this paper suggests a hybrid deep ensemble model that involves the use of the ResNet101 as a deep feature extractor with the Random Forest-based Gradient Boosting (RFGB) classifier as a powerful predictor. Significant features are extracted by the ResNet101 and then they are processed by the RFGB ensemble that combines the strengths of both Random Forest and Gradient Boosting to achieve higher classification accuracy. The proposed ResNet101-RFGB model attained 96% accuracy as compared to the traditional classifiers like Navie-Bayes and Decision Tree (DT), which achieved an accuracy of 88% and 92% respectively. This shows that it has better performance in managing multidimensional data and in improving the accuracy of diagnosis and this means that ensemble learning to a large extent can enhance the dependability and scalability of cardiovascular disease diagnostic systems.

As a Deep Learning approach for detecting and tracking Tapanuli orangutan individuals from thermal infrared footage captured through the use of drones, we developed the improved sparse convolutional and loss-based analysis for drone-based animal movement detection (ISCFLA) model. The detection algorithm analyzes portions of video to affirm the species identification with a 90.25% accuracy upon confirmation by three specialized coders. A Convolutional Neural Network (CNN) approach was used for detection efficiency using annotated orthomosaics as inputs to the computational method. The CNN approach integrated an F1 score classification. A Linear Mixed Model (LMM) approach was used to test differences in the accuracy measures amongst temperature variation. A compound loss metric, sparse convolutional enhanced feature extraction module was developed to improve classification and localization. The revised loss metric function emphasizes difficult samples (Focal Loss (FL)) while the Complete Intersection over Union (CIoU) loss metric improved spatial accuracy of the tracked or detected objects. A Sparse Convolution Enhanced Feature Extraction Module improved effective portions of images, and reduced unnecessary computations and memory usage. The tracker algorithm used in this model ranked detections to ensure consistent tracking despite uncertain detections through the use of a Visual Particle Filter (VPF) utilizing FN on previous locations of particles initialized with bounding boxing center locations. The system computed features Local Binary Pattern (LBP) features within the Region of Interest (ROI) at 4 scales to return a 36-dimensional feature vector for effective target threat identification.

Subtle emotion recognition is a challenging task because low-intensity and ambiguous facial expressions are often overlooked by conventional models. This paper proposes an innovative neural network architecture designed to enhance the detection and classification of subtle emotional cues. The framework integrates convolutional layers for hierarchical feature extraction, pooling operations for dimensionality reduction, fully connected layers for classification, and a cognitive bias-inspired attention mechanism that emphasizes psychologically meaningful micro-features. Unlike traditional convolutional neural networks (CNNs) and attentionbased long short-term memory (LSTM) models, the proposed architecture incorporates contextual priors that mimic human perceptual biases, enabling it to focus on subtle emotional signals. Experiments conducted on the AffectNet dataset demonstrate an accuracy of 94.5% and sensitivity of 83.2%, representing improvements of 9% and 18%, respectively, compared to baseline CNN and LSTM approaches. These results validate the model's capacity to capture nuanced emotional expressions while maintaining real-time feasibility. The proposed system has strong potential for applications in mental health monitoring, adaptive learning technologies, and interactive human-computer systems.

Religious tourism has long played a significant role in India's cultural and economic landscape. Ayodhya, one of the most sacred cities in Hinduism, has recently undergone a transformative development phase, attracting millions of pilgrims and tourists from across the country and abroad. This paper explores the economic transformation of Ayodhya through the lens of religious tourism, examining the surge in infrastructure development, employment opportunities, entrepreneurship, and local business ecosystems triggered by spiritual travel. Using a mixed-method approach with secondary data, government reports, and stakeholder interviews, the study identifies key success factors in Ayodhya’s religious tourism model — including public-private partnerships, integrated transportation systems, cultural preservation, and targeted government initiatives. The findings are then contextualized to Telangana’s tourism landscape, with a focus on religious and spiritual sites such as Yadagirigutta, Bhadrachalam, and Medaram. The paper proposes strategic recommendations that Telangana can adopt to harness the potential of its own religious tourism circuits, including community-based tourism models, digital promotion, sustainable infrastructure, and youth-led entrepreneurship. By drawing parallels between Ayodhya’s transformation and Telangana’s tourism goals outlined in the Telangana Tourism Policy 2023–2030, this study offers a replicable framework for boosting spiritual tourism while ensuring inclusive and sustainable economic growth. This research contributes to the seminar’s theme by providing insights into how religious tourism can be a catalyst for entrepreneurship, rural development, and cultural preservation, and it emphasizes policy-level interventions that can empower local communities and stakeholders in Telangana

Data Hiding is now a major area of study due to the internet’s and multimedia technologies’ amazing growth. Since data digitization and communication networking have advanced over the past ten years, protecting data transmitted over the internet has become a major concern. Numerous applications in steganography and watermarking find it desirable when data is integrated into multimedia cover content, such as audio, video, and photos. Watermarking will be used for copyright protection, while steganography will be chosen for encrypted communication. Even when the embedded data has been extracted, irreparable harm to the cover media is usually unavoidable with such techniques. The perfect restoration of the cover medium is necessary for some applications, including remote sensing, military mapping, geographic mapping, and medical diagnostics. Therefore, in sensitive applications like medical diagnosis, reversible data hiding techniques are used to restore the cover media without distortion. Confidential information, the content owner’s identity, cover content details, etc., are examples of the data that may be included. The data hiding techniques will be shown in order to determine the trade-off between embedding capacity and visual quality. The frequency coefficients of the cover image are changed by the frequency domain techniques. According to experimental results, the recommended procedures perform better than the current approaches for standard QCIF encoded videos and standard grayscale images utilizing the USC SIPI image database. The suggested method considers the information of the cover content during data embedding and provides a trade-off between visual quality and embedding capacity.