• Cleaned the 4 million rows and 37 features dataset using R; Summarized the statistical information of raw data using summary() function; Count 5% missing values and removed rows that have NAs; Used 90 days not repaying the money to define “default” as the label, i.e. 14% default rate (positive label); Generated a cleaned data frame saved in csv file.• Conducted EDA (exploratory data analysis), using box plots and stacked histograms to analyze and compare the distribution of default and… Show more • Cleaned the 4 million rows and 37 features dataset using R; Summarized the statistical information of raw data using summary() function; Count 5% missing values and removed rows that have NAs; Used 90 days not repaying the money to define “default” as the label, i.e. 14% default rate (positive label); Generated a cleaned data frame saved in csv file.• Conducted EDA (exploratory data analysis), using box plots and stacked histograms to analyze and compare the distribution of default and non-default data; Summarized the results in slides and use it as the reference for model building.• Built up logistic regression model to conduct feature selection based on feature importance; Selected 3 most effective factors (age, income, education level) based on the p-value (0.05) given by the logistic regression output; Split training set and testing set to 80/20, rebuild the model using 3 selected features and 5-fold cross validation; Selected 0.7 as threshold based on ROC curve reached 85% accuracy in testing set; Summarized the model performance and feature insights in the ppt report. Show less

• Assist senior business manager to maintain relationship with foreign clients, informed them the latest product information before each exhibition.• Integrated the product lists, make it more efficient and effective.