Qingchuan Lyu

Working at Sphere Institute (Acumen, LLC.)
Studied at Duke University
Living in United States
Website https://sites.google.com/view/QLprofessional

About

I am passionate about building predictive models to deliver new and impactful products. I approach problems with a growth mindset and actively seek new opportunities that push me to constantly learn and improve. My strong technical background, ability to learn quickly, and solid statistics fundamentals have allowed me to successfully move between data mining and machine learning projects.
Contact: Lvqingchuan@gmail.com

Experience
Data and Policy Analyst II - Statistical Programmer
Sphere Institute (Acumen, LLC.)
—Overview: generates statistical results by researching Medicare claims to help the Centers for Medicare & Medicaid Services (CMS) investigate program integrity initiatives

—FISS Pricing Project: achieved 95% pricing accuracy by designing algorithms from scratch according to the most recent policies

—Barriers to Entries Project: identified 2 fraudulent Medicare providers out of over 2,000 Florida practitioners by summarizing billing trends and historical Medicare exemption records

—Ad Hoc Requests: suggested revoking 1 suspicious provider's enrollment for a potential savings of $1.1m in fraudulent billing for Quarter 3, 2019; leads data exploration of urgent ad hoc requests twice a month

—Communication: presents statistical results to clients in bi-weekly calls; explains technical details to policy researchers
Projects
Walmart Food Sales Prediction (Python)
--Investigated the impacts of holidays on daily sales during different times of a year; explored sales trends in different states, item categories, and specific stores across a year"
--Used Dicky-Fuller Test to decide the order of differencing and ACF/PACF plots to decide the order of
Moving-average/Autoregressive models; verified the selected orders by p-values
--Trained Decision Tree, Random Forest and Light Gradient Boosting Tree; achieved 0.25 mean-absolute-percentage-error (MAPE)

Iowa House Price Forecasting (Python)
--Checked if missing values are missing at random and imputed/dropped missing values accordingly; detected outliers and transformed skewed features/labels
--Applied one-hot-encoders and PCA for linear models; achieved 0.11 root-mean-square-error (RMSE) with 5-fold cross validated Elastic Net
--Applied label-encoders for tree algorithms and achieved 0.06 root-mean-square-error (RMSE) with 10-fold cross validated Random Forest