Project Information

Project Name: Drug Discovery using Machine Learning

Project Teacher: S M Hasan Mahmud

Project Summary: Identification of drug target interactions (DTIs) is highly important task in genomic drug discovery and design process. To identify the interaction between drug compound and protein sequence is time-consuming and costly by experimental determination. To overcome this issue, it is necessary to develop effective computational methods for the accurate DTIs identification, which can be used as a lead supporting information in the lab based experimental interaction. In this research, we developed a machine learning based computational method, namely IDTI-ml (identifications drug-target interaction with Machine learning) to identification DTIs using drug structure and protein sequence information. Our method combines Position-Specific Scoring Matrix (PSSM) descriptor and molecular substructure fingerprints. Specifically, the drug molecules are represented by the fingerprints feature to describe the chemical substructure information and the protein sequence is encoded as PSSM descriptor which contains the evolutionary information.

Project Name: Software fault detection using Artificial Neural network

Project Teacher: Hosney Jahan

Project Summary: Regression testing is the task of retesting a software system after changes have occurred, e.g., after a new version has been developed. Usually, only a subset of test cases is executed for a particular version due to restricted resources. This poses the problem of identifying important test cases for testing. Regression testing techniques such as test case prioritization have been introduced to guide the testing process. In our research, we introduce a novel technique for test case prioritization for regression testing based on supervised machine learning. Our approach considers an analysis of the program source code and extracts necessary information, such as software modification history for test case prioritization. We use the machine learning algorithm Artificial Neural Network to evaluate our approach by means of two subject systems and measure the prioritization quality. Our results imply that our technique improves the failure detection rate significantly compared to a random order.

Project Name: Data Steaming (Concept Drift Detection)

Project Teacher: Md. Alamgir Kabir

Project Summary: Concept drift is an important problem in data stream Learning. As data stream is a non-stationary and dynamically changing environment, where the distribution of data can change over time. Therefore, historical data may become inappropriate/outdated and change in data significantly degrade the performance of the learning algorithm, which was built on old data. Concept drift detection and adaptation is very important for learning algorithm. Also, a challenging issue in handling concept drift in data stream is to distinguish between actual concept drift and noise.