Projects

Research Projects:

1. On the Susceptibility and Robustness of Time Series Models through Adversarial Attack and Defense
   
   Team: Asadullah Hill Galib and Bidhan Bashyal.
   
   This research project investigates the vulnerability and robustness of several time series models through adversarial attacks and defense. Experiments are run on seven time series models with three adversarial attacks and one adversarial defense. According to the findings, all models, particularly GRU and RNN, appear to be vulnerable. LSTM and GRU also have better defense recovery. FGSM exceeds the competitors in terms of attacks. PGD attacks are more difficult to recover from than other sorts of attacks. [Manuscript] [Code]
2. Pre-birth Factors in the Early Assessment of Child Mortality using Machine Learning Techniques
   
   Team: Asadullah Hill Galib, Nadia Nahar, and B M Mainul Hossain.
   
   This research project analyzes pre-birth factors, such as birth history, maternal history, reproduction history, socio-economic condition, etc. for the early classification of child mortality. Results show that the it achieved an AUC score of 0.947 in classifying child mortality which outperformed the clinical standards. [Manuscript]
3. Predicting GitHub Issues Lifetime using Machine Learning and Topic Modeling (LDA)
   
   Team: Syed Fatiul Huq, Asadullah Hill Galib, Abu Rafe Md. Jamil, and Ahmedul Kabir.
   
   This research project analyzes the characteristics and applicability of topic modeling (LDA) in GitHub Issues and predicting lifetime based on topic modeling (LDA) with machine learning techniques. It is observed that issues from a large collection of projects can yield distinguishable and comprehensible topics. In terms of predictive performance, the prediction model with topic modeling performs better than the previous approach, with a high increase in precision and f1- measure. Evaluating these findings helps establish topic modeling as a viable feature in issue-based software development processes. [Manuscript]
4. Optimizing Search Space in Code Smells Detection using a Novel Metric
   
   Team: Abu Rafe Md. Jamil, Asadullah Hill Galib, Md Nurul Ahad Tawhid, and Nadia Nahar.
   
   In this study, a simple search space reduction approach is proposed for code smell detection based on a novel software evolution metric of change history information. The proposed approach is evaluated on 11 popular and large-scale projects from GitHub using code smells dataset of four code smells - Blob, Feature Envy, Divergent Change, Parallel Inheritance. Primarily, these four code smells are selected to explore the applicability of the proposed search space reduction approach. The results have shown that the proposed metric significantly reduces the search space while detecting a sound percentage of the actual code smell. It is also analyzed that this approach performs considerably better in detecting Blob, Feature Envy, and Divergent Change, while depicting relatively poor performance for Parallel Inheritance. [Manuscript]
5. ProtectMe: An Approach to Provide Audio Privacy in Real-time
   
   Team: Nishat Tasnim Niloy, Abu Rafe Md Jamil, Asadullah Hill Galib, Naushin Nower, and Mohammad Shoyaib.
   This research project ensures audio communication with privacy in real-time using a modification algorithm. In this study, an approach has been proposed to provide the participants of an audio communication full audio privacy in real time. It will completely hide the vocal identity without compromising any information of the audio stream. This approach does not involve any extra hardware mechanism or data analysis to provide the service. In either way, it is very much user-friendly and inexpensive solution for the users to preserve their security in audio conversation.[Manuscript]
   
   

Academic/Professional Projects:

1. Image-to-Image Translation using Conditional GAN
   The goal of this project is to generate colored images from sketches using a generative model. Conditional GAN-based architecture is incorporated to accomplish the goal. Basically, this project is motivated by the renowned CVPR 2017 paper – Image-to-image translation with conditional adversarial networks (Isola et al.). The architecture and guidelines suggested by the the paper is incorporated into this project. [Manuscript] [Code]
   
   

2. LifeBlood
   “LifeBlood” is an android app for simplifying blood donation system. It is an GPS based blood donor finder android app which sorts out nearer blood donors across the map. In addition, a user rating system and profiling of donors are being implemented. [Technical Report][Code]
   
   
3. Heart Disease Prediction and Factors Analysis
   This project aims at predicting heart disease effectively with consideration of performance measures and analyzing significant factors/attributes. So, it addresses two aspects: how the factors influence heart disease prediction and how well we can predict heart disease. By doing so, heart disease can be analyzed and predicted more effectively.
   Several machine learning techniques with various configurations are employed here for prediction and important factors analysis. UCI repository: Cleveland database is used here for evaluation. Finally, this study suggests an analysis of important factors and recommend machine learning techniques to effectively predict heart disease. [Manuscript][Code]
   
   
4. Analyzing co-authorship network: Centrality Measure, Link Prediction, and Community Detection
   This project aims at analyzing a network of co-authorship relations between researchers (i.e., those in the top-300 highest hindex according to Google scholar and have collaborated at least once with each other). It includes network creation from XML data, data exploration, centrality measure using degree centrality, eigenvalue centrality, and local clustering coefficient. Also, it predicts missing links from the original network using Logistic Regression. Finally, it detects a set of connected components (communities) using a K-means clustering based community detection algorithm. [Code]
   
   
5. AutoPilot-Web
   A web-based digital transformation of BTS (Base transceiver station) management. Its purpose is to optimize and automate the existing network management system. . Its purpose is to optimize the existing network management system. BTS management issues impact revenue generating activities and increase operational costs. As networks expand in size and complexity, availability and performance can be negatively impacted. An optimized and automated network management is a key element to reduce cost and increase agility. Technology used in this project: ✓ Laravel Framework ✓ PHP & MySQL ✓ XAMPP ✓ SourceTree ✓ Postman. [Code]
   
   
6. AutoPilot-Mobile
   A mobile application (iOS and Android) for the digital transformation of BTS (Base transceiver station) management. Technology used in this project: ✓ React Native ✓ React Redux ✓ Redux-Saga (middleware) ✓ Android Emulator ✓ Postman ✓ SourceTree. [Code]
   
   
7. LogBook
   ‘Logbook’ is a web based application which serves as a personal task management tool. This is a simple note taking application that allows you to create text notes, lists, and more. There is are two interfaces- admin interface and user interface. Users have to authenticate to the system to perform tasks. They can add, remove, update notes, lists, tasks, to-do-list, reminder. They can update their information in the account. There is a common introduction page for all the users. From there they can sign up and log in. A non-user can register to be a user. There is a search-bar to search specific note by using heading or tag words. [Technical Report][Code]
   
   
8. Offline-Search-Engine
   An Offline (file-folder) search engine for Linux and Windows operating systems. This project aims to build an optimizing time efficient offline search engine for Linux or Windows operating system respectively. We often face difficulties to use default operating system’s search engine. Like as, it takes too much time for searching. Sometimes we could not find our desired files. This tool mainly concentrates on overcoming that time-wasting process with a little loss of memory. [Code]