Hello!
Welcome to my portfolio website, where data science meets innovation and creativity. Here, you will discover the power of data and its potential to transform the world we live in. My passion for data science drives me to explore new and exciting projects that push the boundaries of what’s possible. From predictive modeling to data visualization, each project showcases my expertise and commitment to delivering valuable insights and impactful results. So, buckle up and join me on this journey of discovery, where data science and imagination come together to create wonders.
Welcome to my portfolio website!
As a data science professional, I am passionate about uncovering insights and solving complex problems through the power of data. With extensive experience in this industry, I have honed my skills in data analysis, machine learning, and visualization to drive meaningful impact and deliver results.
On this website, you will find some projects and experience that showcase my expertise and demonstrate the impact. Whether you are looking to learn more about my background and skills or to collaborate on a new project, this website is designed to provide you with all the information you need.
Feel free to take a look around and get in touch if you have any questions or would like to connect. Thank you for visiting!
Academic Projects
Image Synthesis using GANs
This project focuses on learning the mapping between input and output images in image-to-image translation. The results are optimized using adversarial loss and evaluated using LPIPS Score, IS, and CIS*. The proposed method involves implementing CycleGAN, refining with UNIT GAN, and training BiCycleGAN. The results show that BicycleGAN generates the best images with the best LPIPS, and the project has potential applications in computer vision and has future work in extending to additional domains like videos and text.
Support Ticket classification using NLP
This project aims to deploy a machine learning solution in the IT Service Management (ITSM) environment to categorize service requests. The ML solution uses a supervised classification algorithm and employs Python, RESTful API framework, Scikit-Learn, and SpaCy. The final model has an accuracy rate of 85% and has resulted in significant cost savings and improved SLA response times. The data gathering and exploration phase showed a significant class imbalance issue, which was addressed by eliminating categories with fewer than 100 tickets.
Ames Housing Data and Kaggle Challenge
The project aims to investigate the relationship between various features of a house and its sale price in Ames, Iowa, USA. Using a comprehensive housing dataset, the project fitted different linear regression models to determine the most influential and least influential features on house price. The performance of the models was evaluated using the R2 metric. The findings of the model show that some factors that can negatively affect house value are a damaged garage, having a second floor, and an older house, while increasing square footage, renovating the kitchen and garage, and adding a fireplace could potentially increase value. The model has its limitations as it is specific to houses in Ames, has a small time frame of four years, and may not capture all the factors that affect house price. The model is intended to act as a guideline rather than a perfect predictor of house price.
Life Expentancy and Health Expenditure Analysis in R
The study aims to examine the relationship between health quality and health expenditure in 39 different countries from 2010 to 2015. The data is collected from the OECD and WHO websites and analyzed using 20 different variables. Life expectancy is used as a measure of health quality, while health expenditure in USD is used as the dependent variable for the health expenditure model. The dataset includes factors such as hospital employment, medical graduates, fatalities caused by diseases, and economic factors like expenditure per capita. The data is cleaned and scaled with respect to the population of the respective country, and visualizations are made using R and Power BI.
Research Publication
Blockchain based security for Super-Peer Wireless Sensor Networks
WSN finds its application in a plethora of fields ranging from health care, home application, transport to forest monitoring, surveillance, and military as they are robust and can withstand harsh natural conditions. As the sensor nodes presently work on limited energy and computation power, the ways in which the system can be protected gets limited to a certain spectrum. This paper proposes an innovative decentralized approach for authentication in a super-peer network by the means of the blockchain. By allowing the information to be distributed but not altered, the blockchain has created the backbone for a secure decentralized network. Finally, the proposed system’s performance will be checked by implementing the protocol in the AVISPA tool and checking the performance against various backend tools provided by the tool.