Data Scientist Specialist

Shufan Yu

**** ******** *** *** ****, Clayton, MO, 63105 530-***-**** adxe8z@r.postjobfree.com

EDUCATION

MS, Information System Management Washington University in Saint Louis 2021-Dec 2022 BA, Computational Communication University of California, Davis 2014-2019 TECHNICAL SKILLS

Machine Learning: Classification, Regression, Bayesian, K-means, SVM, XGBoost, Dimension Reduction, Natural Language Processing, Deep Learning, Decision Tree, Random Forest, KNN, PCA, Clustering, Grid Search, Cross Validation Analytical Skills: Data Visualization, A/B Testing, Experimental Design, Hypothesis Testing, Model Diagnosis, Time Series Analysis, Data Mining, Market Segmentation

Data Analysis Tools: MySQL, R, Python (NumPy, Pandas, Matplotlib, Seaborn, sklearn, SciPy), SAS, Anaconda, Google Analytics, Tableau, Microsoft (Excel, Power BI)

H2O.ai, Snowflake, Docker, Spark, Tableau, AWS, Alteryx, SAP, SAS, SPSS, Azure, Google Analytics WORKING EXPERIENCE

Data Scientist @ Firestone Co., Ltd. 06/2022 – 09/2022 Pet Care Product Analysis

• Built Time Series models (ARIMA, XGBoost, Random Forest and LSTM) to predict sales of pet care products using last 5 years’ sales and price data in Python, achieved 847.2 RMSE

• Conducted in-depth product and competitor analysis. Held weekly meetings with business team to present the analysis report.

• Generated 10+ dashboards of insights behind the models including the cause for the seemingly decreasing market share of Revolution®, presented to managers and marketing team monthly using Tableau Risk Order Prevention

• Built a classifier (Logistic Regression, KNN, SVM, Naïve Bayes) with Ensemble Learning (Stacking) to detect high-risk orders

(collusions between salesperson and clinics where orders will be either canceled or returned), achieved 0.927 recall.

• Transformed, cleaned, and conducted Exploratory Data Analysis (EDA) on 3 million sales data using PySpark

• Applied SHAP to find the most important features in ML model, provide recommendations to sales team, reduced 7% high- risk orders in 3 months.

•

Data Scientist @ Kylin Holdings, Co., Ltd. 03/2021 – 08/2021 Fraud Detection for Credit Applications

• Processed imbalanced data with oversampling, undersampling and SMOTETomek. Selected and created features after Data Visualization in Tableau.

• Built and developed a Fraud Detection model with RF, GBDT and XGBoost, achieved 0.913 recall. By Using the ML models, default rate decreased by 5.3%

Customer Behavior and User Experience Analysis

• Monitored and visualized customer engagement and conducted AARRR conversion funnel analysis with user activity data

• Conducted A/B test on experimental and control groups on 480 thousand users for 5 weeks to test correlation between notification push frequency and notification open rate. Calculated the statistical correlation between notification push strategy and conversion rate

• Reported and provided suggestions for stakeholders to increase pushing frequency, decreased user churn rate from -0.38% to - 0.24% and increased conversion rate by 17%

Data Specialist @ Nokia Corporation, Co., Ltd. 05/2020 - 09/2020

• Managed databases and extracted data from large datasets using MySQL and optimized query performance

• Monitored and evaluated key performance indicators for multiple business sectors, altered and replaced indicators based on market changes and business strategies

• Prepared daily, weekly, and monthly report with Jupiter Notebook and MS Excel. Provided Root Cause Analysis for customer satisfaction and sales performance

• Collaborated across departments for indicator effectiveness and maintained a healthy balance between stakeholders SELECTED PROJECTS

Disease Classification and Medical Transcript Analysis 08/2021 – 12/2021

• Analyzed and processed imbalanced data with random oversampling, Encoded text after removing stop word using n-gram bag of words and TF-IDF.

• Build model (Naïve Bayes, KNN and Random Forest) to classify disease ICD category.

• Optimized hyper-parameters with Grid Search and enabled voting ensemble by hard voting, which improved macro average recall rate to 0.873 and f-1 score to 0.802

• Led team collaboration and implemented Agile framework, organized sprint grooming and retrospectives

Contact this candidate