Tianyu Qi

Master's

Data Analytics

Machine Learning Predicting the Accord Sale University of California Berkeley

- Used macroeconomic indicators such as unemployment rates, CPI energy, CPI all, Accord search queries, and miles traveled to predict Honda Accord sales from 2014 to 2023 using OLS regression model - Performed feature importance analysis and multicollinearity checks using Variance Inflation Factor, refining the model by removing high-VIF variables to improve accuracy - Achieved a training set R² of 0.787 and an Out-of-Sample R² of 0.815 on the test data by incorporating additional factors like gasoline prices and month dummy variables - Data Science -Statistical Modeling, University of California, Berkeley 2024.09 - Developed a suite of machine learning models, including classification trees, to segment customers based on credit behavior and identify high-risk profiles for targeted interventions - Conducted stochastic process modeling to simulate customer repayment patterns under varying economic conditions, helping to forecast potential risk exposure scenarios - Built a time series forecasting model using ARIMA and exponential smoothing to predict customer borrowing trends, achieving an R² of 0.82 on out-of-sample data and providing insights for credit risk strategy adjustments