Yangxuan Xu

Experience

• 

  Western University

  May 2022 - Feb 2023

  Research Assistant

  - Facing high expert pricing costs and limited budgets in large actuarial datasets, semi-supervised active learning methods achieved excellent predictions at low costs and the key is selecting the most representative small training set. Our innovative approach used random forest's out-of-bag sample to predict bias, measuring sample ambiguities. Starting from a random small training set, we iteratively add the batch with the largest predicted bias until the budget is exhausted. This method significantly improved efficiency and performance.- Supervisors: Dr. Hyukjun (Jay) Gweon & Dr. Shu Li Show less

• 

  Wuhan University

  May 2023 - Jun 2023

  Research Assistant

  - Worked on the datasets related to Health Geography and Spatial Epidemiology- Supervisors: Dr. Peng Jia & Dr. Yuanyuan Shi

• 

  JD.COM

  May 2024 - Aug 2024

  Data Scientist

  - Employed tree-based machine learning algorithms, including XGBoosting, LightGBM, and Random Forest, to tackle the challenge of predicting rare labels in imbalanced datasets; engaged in variable exploration and optimize data structures to enhance user experience across the complete service journey while conducting business sentiment analysis to proactively identify and address tail risk concerns- Applied unsupervised learning techniques including LOF, Isolation Forest, and AutoEncoder to detect anomalies in claim amounts; by carefully managing service contexts and other potential influencing variables, established a comprehensive monitoring and assessment framework for individual cases and customer service representatives Show less