58 Hiring: Resume Activeness Prediction
The goal is to predicte resume activeness in near future, in online recruitment scenarioes. The project is my thesis and codes here are the experiments.
According to the scenario characteristics, the features extracted from data cover resume user normal information, user activeness history, as well as resume-position matching degree information.
Tested predictors include Logistic Regression, Decision Tree, Gradient Boost Tree, Random Forest, Support Vector Machine, and several hybrid models. The proposed new model RFLRP has good effectiveness in predicting and meanwhile can control the recall rate compared to precision through a single parameter.