jolly-io / jolly-io.github.io

Data Scientist

Technical Skills: Python, R, SQL, AWS, Econometrics, Tableau and Excel Optimization

Education

• M.S., Business Analytics | California State Polytechnic University, Pomona, CA (2023)

• B.A., Political Economics|University of California, Berkeley, CA (2020)

• Unofficial Transcript

Research Projects

Natural Language Processing (NLP) on Microsoft Azure Reviews

In this project, I leveraged the Latent Dirichlet Allocation (LDA) Topic Modeling technique to implement trends analysis on the textual data aggregated from a high impact reviews platform, capterra.com to uncover key trends and insights that can drive customer satisfaction and inform business growth strategies.

In NLP, LDA is employed to understand the main topics in large volumes of text, making it easier to detect patterns, themes and categorize documents based on their content. Given a set of documents, LDA tries to determine the mix of topics that each document represents and the mix of words that define each topic. LDA is a powerful tool in the NLP toolkit for uncovering the hidden thematic structure in a large collection of textual data, which can be impractical to manually analyze for many organizations but bears invaluable insights that inform products innovations, revenue and business growth.

Learn more about this implementation on my Github and the Report

Exploratory Data Analysis on United States Domestic Flights

In this project, I undertook Exploratory Data Analysis (EDA). Given the dataset, I formulated and posed specific insight eliciting questions and endeavored to answer these questions from the dataset leveraging the Pandas and Numpy libraries and relevant Visualization toolkits.

View the implementation on my Github and read the written Report

Telecommunications: Customer Retention and Churn Prediction

In this project, I built a classification model to predict a customer’s likelihood to churn for ZQ, a telecommunications company. Applying the data mining processes, I make the determination among two models (Logistic Regression and Classification and Regression Trees (CART) which was the more effective model to deploy. On the basis of model performance, I proceeded to deploy the Logistic Regression model for the churn classification problem. And conclude with technical and managerial insights for decision makers.

More details on source code implementation here and the formal Report