Brain-computer interface applications in customer experience: Secure data management Shailendra Kumar Rai, Prathwini, Dilora Abdurakhimova, Hameed Hassan Khalaf, Israa Abed Jawad, Pankaj Dadheech Brain Computer Interfaces and Applications in Business, 2025 By providing new and unprecedented insights into consumer behavior and preferences, Brain-Computer Interface (BCI) devices are revolutionizing customer experience management. Brain-Computer Interfaces (BCIs) can accurately anticipate consumer demands, expedite service delivery, and customize interactions with clients by utilizing brain data in real-time. This abstract investigates the potential to enhance customer satisfaction, loyalty, and overall engagement through the integration of Brain-Computer Interfaces (BCIs) into customer experience management systems. BCIs facilitate the creation of customized experiences that can dynamically alter to accommodate individual preferences by conducting a more thorough examination of consumer mood and behaviors. This investigation not only examines the technological challenges and ethical dilemmas associated with the integration of BCI into customer experience management, but also examines the potential future opportunities and current advancements.
Stress Detection in Social Interactions Using NLP and Machine Learning Prathyakshini, Prathwini Proceedings of the 9th International Conference on Inventive Systems and Control Icisc 2025, 2025 Most social media information is utilized for sentiment analysis, reviews, viewpoints, and impact. The suggested approach looks at the posts and comments to determine a person's level of stress by extending sentiment and evaluating emotions. Large datasets of tweets can be subjected to sentiment analysis with machine learning techniques, deep learning models, and BERT for sentiment categorization. Additionally, Latent Dirichlet Allocation is utilized that gathers word groupings and similar phrases that most accurately represent a set of documents by scanning a collection of documents, identifying patterns of words and phrases amongst them, and gathering these data. These emotions can also be utilized to analyze stress and sadness effectively. Both the BERT model and other machine learning models demonstrate excellent detection accuracy. The insights gained from this research contribute significantly to improving mental health. It is evident from the results that the trained model can identify the emotional state based on social interactions when analyzed using a variety of macro and micro variables.
A Deep Learning Approach to Animal Footprint Classification Sinchana A, Prathwini Proceedings of the International Conference on Intelligent Computing and Control Systems Iciccs 2025, 2025 Animal footprint identification is a crucial tool for studying wildlife, protecting species, and monitoring the environment. A deep learning-based approach is presented for automatic footprint identification and classification, focusing on footprints from various animals, including cows, cats, rabbits, squirrels, and elephants. Convolutional Neural Networks (CNNs) are utilized to recognize distinct characteristics of the footprint, ensuring precise classification. A labeled dataset of preprocessed footprint images is used for training, with image augmentation techniques applied to enhance robustness against variations in size, angle, and lighting conditions. For practical application, a user-friendly web interface is developed to enable real-time footprint identification through image uploads. The model demonstrated high accuracy in rapidly identifying animal footprints after testing, with a validation accuracy of 95.65% and a training accuracy of 97.40%. This system offers an effective, non-intrusive approach for monitoring animal populations, examining migration behaviors, and aiding conservation initiatives, rendering it a useful resource for researchers, conservationists, and wildlife enthusiasts.
Animal Track Detection and Classification Using Vision-Based Deep Models Prathyakshini, Prathwini, Disha S Rao, Vaishnavi Kundeshwara Bhat 2025 IEEE International Conference on Distributed Computing VLSI Electrical Circuits and Robotics Discover 2025 Proceedings, 2025 Animal footprint identification is an essential technique for environmental preservation, conservation, and wildlife monitoring. This research demonstrates a deep learning-based approach to identify and categorize the footprints of several animal species, including raccoons, dogs, horses, bears, and rats. This technique uses strong CNNs with deep learning architectures to extract the characteristics of an animal's footprints. To make the system robust to a variety of real-world scenarios, training is done using a tagged and preprocessed image dataset. The challenging problems of size, angle, and illumination fluctuations are then addressed by applying a range of picture augmentation techniques to the images. A web application is created that makes the system user-friendly and accessible by allowing the user to enter footprint images that the trained model will recognize instantly. The model's train and validation accuracies are 96.57% and 93.59%, respectively. This research provides researchers, environmentalists, and wildlife enthusiasts with an efficient, non- invasive method to monitor animal populations, study migration patterns, and enhance wildlife conservation efforts.
Text-to-Image Generation of Bird Images Using GAN-CLS and MS-GAN Prathwini, Prathyakshini, Shreyas A, Raveend S 3rd IEEE International Conference on Networks Multimedia and Information Technology Nmitcon 2025, 2025 This research describes a text-to-image creation system that generates realistic bird images from textual descriptions using the CUB-200-2011 dataset. The technique combines Conditional Generative Adversarial Networks (CGAN) and Mode-Seeking GANs to improve image quality and variety. The pipeline includes Word2Vec embeddings for text preprocessing, a GAN framework trained over 1000 epochs to generate images, and a Gradio-based interface for real-time user interaction. The quantitative evaluation results show an Inception Score of 1.0568, a standard deviation of 0.0044, and a Frechet Inception Distance of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{2 9 9. 0 6 8 5}$</tex>. Classifier metrics indicate 76.3% accuracy and a 6.5% F1-score. Despite minor mode collapse concerns, the system generates visually coherent images with various colors and shapes, highlighting the effectiveness of integrating GAN-CLS with MS-GAN for fine-grained synthesis tasks. This work lays a foundation for future advancements in multimodal learning and text-to-image generation, with potential applications in automated design, education, and accessibility.
Hair Loss Stage Classifcation using CNN and Transfer Learning Approaches Sakshi R, Prathwini, Prathyakshini, Rashmi N, Archana Praveen Kumar Proceedings of 3rd International Conference on Intelligent Cyber Physical Systems and Internet of Things Icoici 2025, 2025 In the field of dermatology, identifying the exact stage of hair loss is essential for early diagnosis, customized treatment planning, and improved patient results. In this study, the performance of three deep learning methods ResNet, VGG16, and a custom convolutional neural network (CNN) for categorizing scalp photos into seven distinct stages of hair loss is compared. A balanced data set of 1,540 images was preprocessed to create subsets for training, validation, and testing. Based on experimental results, the custom CNN achieved a high accuracy of 97.40%, precision of 0.96, recall of 0.98, and F1 score of 0.97, demonstrating a great balance between precision and generalization. With the maximum accuracy of 99.35%, ResNet showed overfitting, which resulted in a lower recall of 0.89. The domain-specific character of the data set made it difficult for VGG16 to generalize, despite its consistent performance with 96.74% precision. The results show that lightweight bespoke architectures may be more appropriate for particular datasets, providing steady and consistent predictions, even when transfer learning methods can produce high accuracy. In addition to highlighting the potential of enhanced CNN models for practical applications in clinical and telemedicine settings, this study adds to the expanding corpus of research on AI-driven dermatological diagnoses.
Cost-Efficient Prompt Routing in Large Language Model Inference Using BERT-Based Difficulty Prediction Harshith P, Anantha Murthy, Harshitha G M, Prathwini, Keerthi Shetty, Roshan D Suvaris 2025 9th International Conference on Computational System and Information Technology for Sustainable Solutions Csitss 2025, 2025 Large language models (LLMs) like GPT-4, Claude 3.5, and Gemini 2.5 are now widely used in natural language processing tasks. But their high inference cost and computing requirements make it hard to scale them, especially in areas with limited resources such as education, customer support, and interactive AI tools. This work proposes a cost-efficient inference setup that uses semantic embeddings from a BERT-based model, a LightGBM classifier to estimate prompt difficulty, and dynamic routing logic to pick the most appropriate LLM based on a balance between cost and accuracy. The system was tested on 990 prompts taken from OpenBookQA (Easy), GSM8K (Medium), and MMLU (Hard). It achieved a weighted F1 score of 0.89, while cutting average inference cost by over 95% from $12.08 to $0.34 per 1,000 prompts compared to always using high-end models. Both semantic embedding and routing logic were crucial in maint aining good performance at cheap cost, according to ablation stu dies, demonstrating that promptaware routing can be applied to real-world LLM inference systems and scale well. This research also evaluate routing framework end-to-end on a mixed workload of conversational, summarisation and code-generation tasks to measure its impact on latency and throughput. Across 1,200 real-world prompts, the system achieves a 40% reduction in average inference time and routes over 70% of queries to lower-capacity models without dropping overall response quality-maintaining an average human-judged acceptability score within 2% of always using the top-tier model. Finally, we release our implementation as a set of Dockerised microservices, complete with pre-trained difficulty predictors and REST endpoints, to simplify integration into existing LLM pipelines.
Assessing University Students' Mental Health through Ensemble Machine Learning: The Role of Academic, Social, and Lifestyle Factors on Depression, Anxiety, and Future Insecurity Keerthi Shetty, Sanjana, Savitha, Anantha Murthy, Harshitha G M, Prathwini 2025 International Conference on Artificial Intelligence and Data Engineering Aide 2025 Proceedings, 2025 Student mental health is becoming an increasingly important problem at academic institutions around the world. The mental health of university students is examined in connection to a number of demographic, academic, and lifestyle variables. Examining the efficacy of students' stress-relieving behaviours was another aspect of the study. Results point to a considerable relationship between students' mental health issues and their lifestyle choices, the amount of work they have to do in school, and their bad experiences on campus.Utilizing algorithms like Random Forest, Decision Tree, XGBoost, and CatBoost, the research focuses on identifying key academic, social, and lifestyle factors that contribute to mental health challenges. The results highlight the predictive accuracy and robustness of ensemble models, offering actionable insights for student mental health interventions.
Lip Reading Detection Using Convolution Neural Network and Recurrent Neural Network with LSTM Sinchana A J, Prathwini, Sonal Alvisha Dchuna, Anantha Murthy, Harshitha G M, Keerthi Shetty 2025 International Conference on Artificial Intelligence and Data Engineering Aide 2025 Proceedings, 2025 Speech recognition system that has fused computer vision with deep learning to recognize spoken words with excellent accuracy. This model uses Recurrent Neural Networks (RNN) with Long Short-Term Memory (LSTM) units to simulate temporal relationships in the speech sequence and 3D Convolu- tional Neural Networks (3D CNN) to extract spatial and temporal characteristics from video inputs, such as lip movements. Even in challenging settings, the capabilities of combining audio-visual information allow the system to collect essential knowledge and ensure that successful recognition is warranted. TensorFlow and Keras libraries are utilized when implementing the suggested architecture, whose performance is marvelous with a 98.5% validation accuracy and 95.7% training accuracy. This technique shows how computer vision and deep learning can be combined to improve the precision and resilience of speech recognition systems.
Harnessing AI for Accurate Bitcoin Price Forecasting: Insights from Historical Data 16th International Conference on Advances in Computing Control and Telecommunication Technologies Act 2025, 2025
Classifying and Predicting Stellar Phenomena using Machine Learning Techniques 16th International Conference on Advances in Computing Control and Telecommunication Technologies Act 2025, 2025
Handwritten Tulu Character Recognition and Translation to Kannada: A Deep Learning Approach for Regional Language Digitization 16th International Conference on Advances in Computing Control and Telecommunication Technologies Act 2025, 2025
Summarization of News Articles Using Deep Learning Techniques Prathyakshini, Prathwini, Ujwal Hegde, Prathviraj Bhat, Kshithij Shetty, Dishan Proceedings of 10th International Conference on Communication and Electronics Systems Icces 2025, 2025
Autism Prediction using Machine Learning Keerthi Shetty, Savitha Savitha, Sanjana, Anantha Murthy, Harshitha G M, Prathwini Prathwini Proceedings of 6th International Conference on Intelligent Communication Technologies and Virtual Mobile Networks Icicv 2025, 2025
Deep Learning Based Automated Lip Reading for Deaf Prathyakshini, Prathwini, N Pratheeksha Hegde, Vaishali, N Rashmi, Archana Praveen Kumar 2024 3rd International Conference for Advancement in Technology Iconat 2024, 2024
