🎩 Jason Fang Email & Phone Number

College Park, Maryland, United States

▪ Team LeaderMentor fellow data scientists and help them learn business needs, project strategies, and presentation skills. Provide coding and technical support if necessary.▪ Time Series for Budget Allocation Collaborate with the Direct-to-Customer team to allocate a $300 million budget across marketing channels. Build ARIMA and Random Forest models to predict the next quarter’s sales. Utilize spend optimizers to maximize sales based on the models. Increase sales by 3% on average every quarter. ▪ Ensemble Models for Customer Acquisition CostCombine the random forest feature importance and linear regression coefficients to calculate customer acquisition cost. Compare the cost with the attribution cost and spend curve to provide important insights to stakeholders. ▪ Causal Inference for A/B Testing Build propensity models to review the effectiveness of marketing channels. Replace traditional A/B testing experiments with faster approaches. Evaluate each country’s performance and help the company to expand its global markets in more than 20 countries in Europe, Asia, and so on.

Washington Dc-Baltimore Area

Visit my magic website: jasonfangmagic.com for more info• Magic Shows• Balloon Animals • Face Paintings

Gaithersburg, Maryland, United States

▪ SQL for Website AnalysisUse SQL/AWS Redshift to derive key website performance metrics, such as traffic source, landing page performance, conversion rate, bounce rate, and so on. Create Tableau dashboards and present the results to stakeholders.▪ Classification for Website ImprovementApply XGBoost, Logistic Regression, Artificial Neural Network to analyze website data, optimize marketing campaigns, and then redesign the website, increase the conversion rate by 2% and decrease the bounce rate by 5%.▪ Multi-armed Bandit for A/B TestingConduct quantitative (historical data) and qualitative analysis (focus group), define the sample size, select independent samples, make win-lost tradeoffs and analyze A/B testing data by using Thompson Sampling/UCB to design the website and newsletters. The new website and newsletters design extended average session time by 12% and increased the click-through rate by 6%.▪ Time Series for Profits/Revenues PredictionUse SARIMAX, Recurrent Neural Network (LSTM), Facebook Prophet, Multiple Linear Regression to analyze seasonal data with exogenous factors. Use models to predict customer numbers in order to arrange staff schedules during holidays and shut down the facility on Tuesdays due to low profits, which resulted in increasing profits by 4%.

Greenbelt, Maryland, United States

▪ Clustering for SegmentationImplemented K-means and Self-Organizing Maps to segment football viewers in China. Analyzed features from different segments and organized Super Bowl watch parties for old and new football fans in Beijing and Shanghai, which attracted more than 50 people in each city.▪ Regression for Football News Pageviews ImprovementApplied linear regression models, Random Forest, and Gradient Boosting to analyze football news pageviews and determined feature importance in ensemble learning models. Found out what players and teams Chinese fans like to read and asked editors to publish more news that can attract more people to know football. Increased 8% pageviews on average.

Master Of Science - Ms, Marketing Analytics

Bachelor Of Science - Bs, Marketing