Machine Learning Engineer
Current2018.1 - present• Developed anti-fraud rules with rule-extraction from CART/ Regularized Random Forest. Achieved a decrease of 20% in transaction disturb rate while keeping same level of recall rate of fraudulent transaction• Feature extraction and derivation from trading behavior, LBS, devices and other data sources to support customer marketing and application scorecard development (XGboost) for personal loan product• Understood offline collection business scenario, analyzed user behavior and built customer segmentation model with K-means++ to help differentiated marketing • Summarized the analytical needs from various departments, conducted data cleansing/manipulation/feature extraction and designed customer label system2017.1-2017.12• P2P merchants risk modeling using Random Forest + SMOTE + Leave-One-Out to overcome sample imbalance• Trained face verification model using OpenFace framework and achieved TAR>97% under FAR<0.01% and the model was applied in both 1-to-1 and 1-to-N scenario2016.1-2016.12• Used CNN to reject fake e-signature and solved generalization problem caused by location/grayscale of different e-signature• Ad-hoc analysis with business departments on topics including customer churn prediction, marketing anti-cheating, abnormal merchant detection and so on 2015.7-2015.12• Automated the operation reports using Python and analyzed the fluctuations in key indicators. Assisted in the writing of 2015 “White Paper on Shanghai Internet Credit Service Industry”