Employee Attrition Prediction with R Accelerator

speaker: Le Zhang, Data Scientist at Microsoft event: FOSSASIA pre-meetup

Client project done by Microsoft Algorithms and Data Science team

Team data science process https://github.com/Azure/Microsoft-TDSP

Identify what features are needed to build the model

• Look at reasons for employees to leave and corresponding features

Data collection

• HR dept
• IT dept
• direct reports
• social media network

Data understanding

• visualization
• basic statistics

Feature engineering

• statistics
• time series model
• text mining

Feature selection

• correlation analysis
• model based feature selection

Model selection and validation

R accelerator

• template with R scripts
• helps to create a proof of concept

Resampling to balance the data classes

• SMOTE