Thao Phan's Projects
Developed a housing price prediction tool for the real estate company to forecast upcoming housing prices by employing a linear regression model with an R-squared value of 70% based on a 13-month dataset of King County residences.
Analyzed data on 200+ studios and 3000+ movies to advise Microsoft on investing in a new movie studio. Recommended Animation genre with a budget of 50-200M USD to double ROI, and managed popularity of 20+ to ensure global revenue greater than 500M USD.
Open source platform for the machine learning lifecycle
NYC Data Science April 24 Cohort
Implemented an image search feature on the website to improve the experience of 35% of customers dissatisfied with keyword searches. This enhancement incorporated the CNN model with 95% accuracy for image classification and the Siamese model for the top 5 similar products, leveraging the data set of 24,000+ images across 10 different classes.
High-Resolution Image Synthesis with Latent Diffusion Models
Reduced the initial churn rate from 25% to less than 20% by implementing churn-prevention alerts based on the predictions of customer churn likelihood built from the XGBoost classification model with 95% accuracy from a 6-month subscriber dataset.
Generated automated weekly sales forecast by departments in 45 stores using a dataset of weekly sales data spanning over 2 years, by employing the Random Forest Regressor model with achieving the Mean Absolute Error (MAE) of less than 10% of the forecast.