Nowadays， employee turnover is an important issue in organizations. Predicting employee turnover in advance using big data can provide a scientific basis for making decisions of employment and retention， which will enhance the foresight and wisdom of human resource management. The key problem of reported employee turnover prediction models is the lack of business driving force， which is manifested in that they can only answer to the human resource manager whether an employee has the turnover intention， but cannot further indicate why he or she will leave and how to retain it in a targeted manner. Therefore， focusing on this problem， a portrait-based employee turnover prediction model named PCC is proposed， integrating PCA （Principal Component Analysis）， CLARA （Clustering Large Application）， and CART （Classification and Regression Tree）. Finally， the PCC model is experimented on an open-source employee turnover dataset with 14，999 samples from Kaggle. Theoretical and experimental studies show that the PCC model can provide theoretical reference for employee turnover prediction， turnover portrait description and accurate retention strategy design. Besides， it can be used to predict the turnover intention of individual employee and monitor the turnover distribution of all employees. Furthermore， it can provide data support for employee retention and talent reserve from both micro and macro levels. In summary， the PCC model is a feasible， effective and intelligent system for human resource management.