About me

I am pursuing a Computer Science-Big Data Track Graduate program at Kennesaw State University since January 2018.

My thesis research is focused on deep learning in Computer Vision, specifically for Cancer detection and classification with digital pathology data. I am a machine learning, data science and data engineering enthusiast with demonstrated projects in Computer Vision, Natural Language Processing and Recommendation engines.

My main research projects are:

1. Deep learning on Carcinoma Cancer detection and classification

   • Digital Pathology analysis using Deep Learning models

   • Leveraging deep learning models on Cervical/Carcinoma Biopsy slides with Tensorflow and Pytorch

   • Visualizing target cancerous tissues and scoring using heat maps

   • Developing end to end ETL pipelines to enhance automatic and accurate decision making by pathologist.

   • Through the CRISP-DM model and ETL pipelines, a Classification score of 96.4 percent is achieved on Adenocarcinoma stomach epithelial cells from whole slide images (In preparation for IEEE BIBM )

2. Identify segments of the population that forms core customer Base in Germany

   • Through unsupervised and supervised learning techniques, the core customer base of a mail-order company is identified.

   • Two real life datasets are leveraged; the first consists of 891,221 people from the demography of Germany, and the second consists of 191,653 customers from the company, each with 85 engineered features

   • Achieve a predictive accuracy of 82 percent after visualizing, cleaning and ingesting the data in defined mathematical models

   • Complete technical and functional Implementation description is free to the public on Medium (https://medium.com/@tsakunelsonz/https-medium-com-tsakunelsonz-a-gentle-customer-segmentation-approach-for-mail-order-industries-a10e3c78e40b?source=friends_link&sk=fac1906f04b83de91f2e27c1e9d26d91)

3. Boosting recommendation engines through an offline testing environment to avoid customer churn

   • The three main categories of recommendation systems are introduced and independently implemented:

     • Knowledge-based recommendation engines free on GitHub (https://github.com/Tsakunelson/knowledgebased_recommendations)

     • Collaborative filtering recommendation engines free on GitHub (https://github.com/Tsakunelson/content_based_recommendation_engine)

     • Model-based recommendation engines free on GitHub (https://github.com/Tsakunelson/Model_based_recommendation_engine)

   • A new strategy that integrates domain knowledge for new users into a machine learning based recommendation system for existing users is adopted

   • An offline/online evaluation strategy is introduced to reduce customer churn

   • This enables the deployment of accurate recommendations for both new and existing users.

   • Using the movie Tweetings dataset, a boosted system is created, which addresses two major issues faced by current recommendation systems: client churn and the cold start problem.

   • Published in the ACMSE 2019 conference (Acceptance rate 45 %)