Area of Interest:
Artificial Intelligence and Algorithm
Machine Learning
Data Mining
Graph mining
Game theory
Deep Learning
Natural Language Processing
Ongoing:
Determining Correlation Between LTE Usage and Regional Welfare in Bangladesh
Under supervision of Dr. Md. Shohrab Hossain, Associate Professor, Dept. of CSE, BUET
In collaboration with Robi Axiata Limited
Prediction of B-factors from Protein Sequence
Under supervision of Dr. M. Sohel Rahman, Professor, Dept. of CSE, BUET
• The objective of this research is to predict normalized B-values in Amino Acid chains, which can be later be used to determine the thermal stability of proteins.
• We have formulated a novel sequence to sequence regression model and implemented using Scikit-Learn. We have our regression model with RNN & LSTM based regression models using PyTorch. So far, this model gives comparable results with existing works. We are currently doing rigorous testing on this model.
Past:
Malware Detection Techniques in Android [2016-2017]
B. Sc. Thesis, under supervision of Dr. Md. Shohrab Hossain, Associate Professor, Dept. of CSE, BUET
• In this research, we had performed dynamic analysis of Android apps by sequencing System calls. From there we had extracted a set of distinctive features based on Kernel call logs using Markov Chain and trained on a Machine Learning model.
• We have validated the effectiveness of our model by performing experiments on a large and hybrid dataset. We combined the malware dataset obtained from Malware Genome Project (http://www.malgenomeproject.org/) and Android Malware Dataset (provided by ArgusLab).
• We have also provided a novel feature selection heuristic, namely Logarithmic Gaussian Dissimilarity (LGD), which is general purpose and can be used in any Machine Learning 2-class classification problem. The heuristic gives an estimation of how effective a feature is; which can hence be used for feature-dimensionality reduction.
• The finding of this research has been accepted for presentation at IEEE Global Communications Conference (GlobeCom) 2018 NGNI and for publication in its proceedings.