Sales Data

YUAN LING

Address: ** ***** ***** *** ****, Jersey City, NJ, 07310 Email: ******@********.*** Tel: 612-***-**** EDUCATION

Columbia University, Graduate School of Arts and Sciences New York, NY Master of Arts in Statistics, GPA: 3.80/4.00 February 2015 University of Minnesota-Twin Cities Minneapolis, MN Bachelor of Arts in Statistics, GPA: 3.74/4.00 May 2013 Qingdao University Qingdao, Shandong, China

Bachelor of Management in Information Management and Information System June 2012 Computer Skills: R, SAS, SQL, VBA, MS Office (Word, Excel, PowerPoint) Statistical Skills: Logistic Regression, Random Forest, SVM, KNN, Decision Trees, Cluster analysis, Bayesian Certifications: SAS Base, SAS Advanced

PROFESSIONAL EXPERIENCE

Class Wish New York, NY

Intern, Data Analyst Spring 2016

Grabbed the data about nonprofit organizations from website by Python Set up an Access database and imported those data into it; used MySQL to manage them.

Identified which nonprofit organizations are valuable by building predictive models using the useful data. International Academic Alliance New York, NY

Intern, Analyst Summer 2015

Managed the students’ account, user access and permissions.

Utilized survey strategies and data analysis to predict which course will be the most popular for the next semester. PROJECT EXPERIENCE

Columbia University, New York, NY

“Valuable Shopper Identification: Predict Which Shoppers Will Repeat Purchase” Fall 2014

Performed data reduction and feature engineering using Python. Reduced the data from 350 million to 1.8 million and created 47 predictors.

Evaluated 6 models including Logistic Regression, SVM, KNN, CART, Random Forest, and Boosted Tree with R. In each model using crossing-validation, bootstrap or adding shrinkage to do some adjustments.

Logistic regression with stepwise was selected as the best model with AUC of 0.5947 and accuracy of 0.7389; the brand-related predictors dominated in each model.

Columbia University, New York, NY

“Predict the time to return of drug use” Fall 2014

Collected a sample from 575 subjects. Merged the 9 predictors by SAS.

Checked the significance of each predictor. Fitted all significant predictors in a multivariable Cox-proportional hazards model, and then used backward elimination to choose the best model.

The ‘best’ set of predictors for modeling time to return of drug use is the set consistent with the Treatment Randomization Assignment, IV Drug Use History at Admission, and the Number of Prior Drug Treatments. Columbia University, New York, NY

“Predict the Residential Home Sales Prices” Fall 2013

Collected data on home sales from a Mid-western city during year 2002.

Created and evaluated 13 different linear and non-linear, tree-based and rule-based predictive models; modified each model using crossing-validation.

Boosted Tree was selected as the best models with RMSE of 55632 and R of 0.859 ACTIVITIES

Content Developer (Online Mentor): Noiz Ivy, Inc, Wading River, NY, United States, 2014 Grader: Department of Statistics at Columbia University (Course: Data Mining), New York, United States, 2014 Volunteer: Sino-American Culture Association’s Annual Conference, New York, United States, 2014

Contact this candidate