• Calculated sentiment score for each member and ranked the top 10 most unhappy members • Found out common keywords for disenrolled members and enrolled members using member enrollment census data and demographic information

Classified if the patients are responders or non-responders by utilizing K-nearest-neighbor (78.6%), Random Forest (82.3%), Simple Neural Network (92%) and Support Vector Machine (79.3%)Used dplyr to make summary statistic table of drug concentration according to different treatment level.Used ggplot2 to determine if MEDI6012 has drug effect on heart rate, if there's no time delay between drug concentration and baseline-corrected QTc interval and whether QTc interval is independent of HR.Implement Linear mixed model to develop the relationship between drug concentration and baseline-corrected QTc interval

Conducted comparison analysis for a vegan recipe generator startup company and an emotion sensor startup company.

Merged manager table and customer table and organized the new table by state in MySQLUsed ggplot2 to determine the trend of the total number of default customers in different states for the past 10 years in RUtilized Random Forest to forecast the total default amount in each state with accuracy rate 82.1% and determine which variable has the most predictive influence in RImplemented Decision Tree to predict whether new customers will default or not with accuracy rate 84.36% in R

• Graded assignments for Introduction to College Mathematics Course.• Answered daily emails from students about course content, assignment and exam grades.

Applied linear regression in scikit-learn to predict daily number of people open new account.