Fast string-similarity algorithms and blocking techniques for linking large, noisy datasets, developed under a U.S. Census Bureau–University of Connecticut cooperative agreement. Includes a linear-time Jaro similarity algorithm, SuperBlocking, and phonetic blocking.
Bit-parallel and GPU algorithms for record linkage that process millions of records and billions of candidate comparisons.
Cluster-quality metrics and local-search optimization for evaluating and refining record-linkage results when true matches are unknown.
Problem-specific communication strategies for distributed quantum algorithms such as the Quantum Fourier Transform and Grover’s search.
Developed an efficient computational tool on CPLM-2.0 database proposing a prediction scheme based on evolutionary conserved information of protein sequences. The study explored the efficacy of protein sequence based features and protein secondary structure for predicting Lysine Succinylation sites using machine learning techniques. (Click on the title for a brief overview)
RNN with LSTM model on DSE data
On Credit Vending Machine for BUET Students, Teachers & Staffs
An easy-to-maintain Information System Design for a School