IEEE-CIS Fraud Detection, TOP 1% Range – 30th of 7416 competitors IEEE Oct 2019 • Analyzed over 1.2 million lines of transaction data primarily using R and Python• Determined 414 missing value in data set using numpy pandas and scikit-learn• Utilized Light GBM and histogram based algorithms to support categorical features and improved efficiency• Completed a decision tree to observe the performance of the test set and fitted the model with num-leaves• Adopted blending methodology in model ensembling and weighted categorical coefficients… Show more • Analyzed over 1.2 million lines of transaction data primarily using R and Python• Determined 414 missing value in data set using numpy pandas and scikit-learn• Utilized Light GBM and histogram based algorithms to support categorical features and improved efficiency• Completed a decision tree to observe the performance of the test set and fitted the model with num-leaves• Adopted blending methodology in model ensembling and weighted categorical coefficients to predict results• Performed feature analysis to obtain the distribution of features on the train set and the test set• Analyzed traits on both positive and negative samples• Evaluated consistency of the distribution of the train set and test set by using Facets Show less