yuchenwei

Built and fine-tuned time series model: SARIMAX model to predict the number of new hires using data pulled from an Oracle server, reducing manual processing time by 75%. Transformed daily passenger data with over 60,000 rows for the bus department, converting original reports into structured CSV files. Implemented Python scripts, enabling automatic daily updates in the Oracle database. Fixed mistakes in a deployed model and presented my findings to two departments. Simplified the model by reducing the BIC value by 80 on average with 10% more accurate predictions.

Developed Machine Learning-based quantitative investment strategies to optimize portfolio performance and manage risk. Invented and Back-tested novel alpha signals to predict the movements of markets over time horizons spanning from minutes to days.

Analyzed Johnson & Johnson’s medication data with over 110,000 rows and built machine learning models including Random Forest Regression, AdaBoost, XGBoost to predict the number of offtake, improving the R-squared by 65%. Quantified the relationship between offtake and sellout for Tylenol by fitting a Linear Regression model, achieving an R-squared above 0.9 for over 2000 chain stores. Provided business insights on categorizing different chain stores and reduced the risk of inventory shortages by 95%.