XinYuan Xu

Bellevue, WA 858-***-**** ad5brf@r.postjobfree.com www.linkedin.com/in/xinyuanxuba2022

EDUCATION

M.S. Business Analytics January 2024

Boston University Questrom School of Business, Boston, MA B.A. International Business August 2022

University of California San Diego, San Diego, CA

PROJECT EXPERIENCE

Sign Language Prediction Model Project April 2023 - May 2023 Questrom School of Business, Boston University, Boston, MA

• Utilized advanced data preprocessing techniques to enhance the quality and accuracy of the dataset, ensuring improved model performance and predictive capabilities

• Explored a variety of model architectures, including dense, Conv2, and CNN, to maximize the predictive performance of the sign language gesture prediction models

• Achieved a 20% increase in model accuracy through extensive hyperparameter tuning and grid search, demonstrating the potential of fine-tuning model settings

• Ensured model interpretability by implementing Lime, providing valuable insights into the decision-making processes of the complex models, ultimately improving the model's trustworthiness and usability Large Climate Data Analysis January 2023 - February 2023 Questrom School of Business, Boston University, Boston, MA

• Processed a vast dataset of 869,313 city responses from CDP's 2020 questionnaire, conducting data preprocessing, one-hot encoding, and aggregation

• Utilized GeoPandas for geographic data visualization, creating polar graphs to illustrate countries' climate adaptability challenges

• Performed geolocation and added latitude and longitude data for precise mapping and location-based analysis SBA Loan Analysis January 2023 - February 2023

Questrom School of Business, Boston University, Boston, MA

• Analyzed 899,164 rows and 27 columns of SBA loan data to predict outcomes under loan dataset context

• Explored various machine learning models (Logistic Regression, Dimension Reduction, LightGBM, XGBoost) within the financial domain, resulting in a diversified approach to risk prediction

• Utilized a correlation heatmap to visualize feature relationships, uncovering insights into data patterns

• Leveraged data visualization techniques to illustrate the relationship between loan terms and loan outcomes

• Developed data preprocessing pipeline, achieving an accuracy score of 99.46% on the test dataset

• Recommended actionable enhancements for commercial banks, resulting in a decrease in loan charge-offs PROFESSIONAL EXPERIENCE

Natient Work Technology LTD., Jiangsu, China August 2020 Data Scientist Intern

• Collaborated with the technical team, made fraud detection projects with customer transaction data, and raised accuracy of predicting the probability of fraud by approximately 20%

• Utilized SQL to complete ETL for the transaction data and then used both SQL and Python (Pandas package) to conduct basic data cleaning, data aggregation and exploratory data visualization to better analyze the data and build models

• Provided advanced data analysis and built different classification models such as logistic regression, random forest, XGBoost, support vector machine classifier to predict the fraud data in both training and testing dataset

• Conducted K-fold validation to tune the hyper-parameters and then chose the best model with the most effective set of hyper- parameters and improved the accuracy from 61.42% to 82.31% in holdout set RELEVANT SKILLS

Technical Skills: SQL, Python, MATLAB, RDBMS, Microsoft Office (Excel, PowerPoint, Word), Spark, GCS, Tableau, Machin learning, Neural Network, Keras model, Language modeling

Relevant Coursework: Business Analytics, Statistics, Product Management, Business Strategy, Neural Network, Advanced Analytics, Machine learning

Languages: Mandarin (Native), English (Fluent), Japanese (Entry)