(Github Repository)
Delegated Access Control using Attribute-Based Encryption
Planned to solve some real-world issues and challenges faced in the EHR system stored on the cloud. Built an application using the Python Django framework. The application uses user attributes to control EHR access. ABE was used for encrypting EHR. An ontology was also built following the HIPAA act to control access to different fields of an EHR. The framework allows search over encrypted data that was built using the searchable encryption techniques.
[Link]Contact Tracing System
Designed a contact tracing system for covid19 using PL/SQL. Started designing the database with an ER diagram, inserted some sample data, and implemented a set of features. Each feature was implemented as one or more Oracle PL/SQL procedures/functions.
[Link]Spam Detection using machine learning approaches
Built predictive models that can be used to identify spam email. Tried to infer the relationship between the predictors and the response.
[Link]Churn prediction for customers in the banking system
Built machine learning and deep learning models to predict customers which are likely to churn. Identified the features that are important for the prediction.
[Link]Predicting movie genre from plot summaries using Support Vector Machine
Built classification models to predict movie genre from plot summaries. Used data pre-processing techniques to pre-process the open source data.
[Link]Is Your Better Half Prone to Divorce? Predicting Divorce Using Data Science
Built machine models to predict divorce using an open-source dataset. Experimented with feature selection and data pre-processing techniques. The accuracy of the model was compared before and after feature selection.
[Link]Integrating information from multiple repositories
Built a layer of integration above the local databases. The integration layer is the layer of metadata that includes information that defines the local databases. Used dynamic SQL to merge different databases. Metadata layer was also used to map the same entities with another name.
[Link]Efficient and Flexible Aggregation and Distribution of MODIS Atmospheric Products
Built models to process the data for a day or month within the shortest possible time. Implemented different parallel processing techniques and re-sampling methods in python to reduce computational time.
[Link]Assessing water budget sensitivity to precipitation using VIC hydrologic model
Used VIC to test the effect of precipitation uncertainties on water budget components for the Potomac river basin from April 2017 to September 2017, that was deployed on taki, UMBC HPCF. Analyzed the monthly water balance components’ sensitivity by increasing variability in input precipitation using parametric resampling methods.
[Link]