Project Report (Bachelor of Science in Computer Science)

Permanent URI for this collectionhttps://hdl.handle.net/10361/28439

Browse

Now showing 1 - 6 of 6

Mortality risk and health suggestions for critical patients using extended LSTM and CNN
(BRAC University, 2025) Ashraf, Muaz Ibne; Islam, S.M. Sihat; Siraz, Zasia Farzin; Rahman, Md. Khalilur; Department of Computer Science and Engineering
Intensive care units (ICUs) and their high mortality rate often require predictive tools that could help to determine at-risk patients in time and direct the interventions. The given project presents a deep learning system that merges Convolutional Neural Networks (CNN) with the Long Short Term Memory (LSTM) networks to better predict the risk of mortality at an early stage among critical patients in ICUs. Using a significant portion of the patient data such as the vital signs and laboratory results, the model can carry out constant risk analysis and it could prove superior to the conventional scoring systems. Health suggestion module is also incorporated to make suggestions on clinical interventions to be made on high risk patients thus assisting healthcare providers with their decision-making. Finally, the suggested direction will enhance the patient outcomes as it will allow delivering proactive medical care and, thus, distributing ICU resources more reasonably.
An accessible mental health app for early depression management called Support Sphere
(BRAC University, 2025-06) Akter, Jahanara; Rahman, Md. Khalilur; Department of Computer Science and Engineering
This research paper presents the design, development, and preliminary evaluation of a mental health support application aimed at individuals experiencing mild to moderate levels of depression, specifically focusing on users in stages 1 and 2 of the condition. The project was conducted as part of the CSE400 Project course requirement for the B.Sc. in Computer Science program at Brac University, as a Project. This application seeks to provide users with a pressure-free, judgment-free space for self-care. The core features of the app include a gratitude journal, breathing workouts, a happiness-boosting to-do list, inspirational content, and progress tracking mechanisms, all built using the Flutter framework with cloud-based data storage. The motivation for the project stems from both personal and professional experiences of the author, who has dealt with Major Depressive Disorder and has previously worked at a suicide prevention helpline. Existing mental health apps often operate under freemium or premium models, making them inaccessible to many users, especially in developing regions. This app offers a free, simplified, and stigmafree alternative. Due to the sensitive nature of mental health, the application was evaluated through qualitative feedback from a small group of users rather than largescale data collection. Their emotional responses, usage behavior, and suggestions helped validate the app’s impact and informed the design adjustments. This paper also explores the societal, ethical, and environmental implications of such technologies and provides an extensive literature review and gap analysis of current mental health apps and studies. Due to the sensitive nature of the concept, the application went through a small testing phase. In general, participants found the application to be helpful, journaling, breathing and BMI modules were particularly praised to help with anxiety and body image.
An analysis of personalized learning platform model
(BRAC University, 2024-10) Rabbi, Golam; Shakil, Arif; Department of Computer Science and Engineering
In the modern continuously developing field of education, it is concluded that the establishment of individual learning appears to be possible with the help of machine learning solutions. The case being presented in this paper calls for the implementation of a concept of a learning platform that is packaged with a state-of-the-art machine learning tool set to boost academic achievement. The proposed method consists of three main components: This paper presents a seven-feature approach that includes detailed response and feedback, dynamic control of learning processes, and a recommendation system. The recommendation system applies information, demographic and collaborative information about learning resources to each learner based on individual learning style and academic accomplishment records. Adaptive learning thus self-organizes content based on the various aspects of student interaction and achievement so that a perfect learning path is achieved. Feedback as motivation and constant improvement, feedback for timely, useful, and individualized criticism through sentiment and Natural language processing. The integration of these components, however, suggests the potential for developing the present recommendation platform to offer a more productive and enjoyable educational experience that meets the needs of every learner. This outsiders’ concept seems to have the potential to outcompete traditional normative pedagogy teaching models, in the sense of efficacy and effectiveness as depicted by learners’ performance and satisfaction.
Optimizing endpoint detection and monitoring in enterprise solution : a cyber threat intelligence approach
(BRAC University, 2024-09) Sarker, Apurba; Mondal, Joty Prokash; Pran, Suzaur Rashid; Islam, AR Rafiu; Seraj, Mehnaz; Department of Computer Science and Engineering
Advanced cyber threat intelligence systems are crucial in a time when enterprise solutions are increasing and are being targeted by more sophisticated cyberattacks. This research aims to study ways to improve endpoint detection and monitoring in an enterprise company by installing a Security Information and Event Management (SIEM) system based on Wazuh with integration into ELK Stack. The report performs an inside-out examination of the integration and deployment capability for each technology, focusing on real-time anomaly detection and threat mitigation toolkits at complied states. Together, these techniques create a powerful combination of analytical security, intrusion detection, log data analysis, file integrity monitoring, and vulnerability management capabilities being adopted in a variety of industries that handle sensitive data. This infrastructure uses the Wazuh active response module to detect security threats and look for indications that one is starting up. Denial of Service (DoS), brute-force attacks, simulations, and integrity file delinquencies with tests as proofs tell stories about a good performance estimation when Elasticsearch and FileBeat application is used jointly. Wazuh provides a robust and cost-effective solution for enhancing the security posture of enterprise solutions. Wazuh instantly detects and monitors simulated attacks such as denial-of-service (DoS) attacks by spotting suspicious file changes in real-time, logging failure authentication attempts, and identifying the root source of the flood of requests. This study also provides useful insights on designing and deploying comprehensive cybersecurity solutions with opensource tools such as Wazuh, making visual insights for file integrity monitoring (FIM) in real time.
Determining intensity of mental state of an unsound individual through text using ML
(BRAC University, 2024-10) Khan, Nishat Sabah; Rahim, Md. Sazidur; Hossain, Muhammad Iqbal; Department of Computer Science and Engineering
This research investigates the application of machine learning to detect and classify the intensity of various mental health conditions through text analysis. By analyzing user-generated statements, the study aims to identify patterns that correspond to different mental health states, such as Anxiety, Depression, Bipolar Disorder, and Suicidal tendencies. Through rigorous text preprocessing and feature extraction methods, meaningful insights are drawn from the data. The performance of the proposed approach is evaluated through standard metrics, demonstrating its potential to support mental health professionals by automating the initial stages of mental health screening. The findings highlight key challenges, such as language complexity and emotional context, and offer directions for future work to enhance the system’s accuracy and adaptability. This research provides a foundation for developing scalable, automated tools that could be integrated into mental health care and online support platforms.
Enhancing optical character recognition capabilities for Bengali script: the development and evaluation of Bengali CharNet
(Brac University, 2023) Kamal, Rizvy Ahmed; Rasel, Annajiat Alim; Department of Computer Science and Engineering
Optical Character Recognition (OCR) technology has made an excellent stride in recent years, yet the accurate digitization of Bengali handwritten script remains a formidable challenge. This project introduces ’Bengali CharNet’, an improved deep learning-based model, specifically designed to advance OCR capabilities for Bengali handwriting, which is notably intricate and diverse in its character composition. The project aims to fill a crucial gap in OCR technology’s effectiveness with complex scripts like Bengali, which is the seventh most-spoken language in the world. The results of this research project are significant, with Bengali CharNet demonstrating a remarkable improvement in accuracy, precision, and recall compared to existing OCR models. The model achieved an overall accuracy of 96.8%, showcasing its effectiveness in recognizing and digitizing Bengali handwritten characters. This achievement represents a substantial advancement in the field of OCR, particularly for scripts that possess a high degree of complexity.

Browse

Recent Submissions