Feature Screening for Interval-Valued Response With Application to Text Mining in Online Job Markets

Text mining of online job advertisements and estimation of return to skills have attracted great interest in the recent research in labor economics. In this paper, we study the relationship between the posted salary and the job requirements in online labor markets. There are two difficulties to deal with. First, the posted salary is always presented in the interval-valued form, for example, 5k-10k yuan per month. Simply taking the mid-point or the lower bound as the alternative for salary may result in biased estimators. Second, the number of the potential skill words as predictors generated from the job advertisements by word segmentations is very high and many of them may not contribute to the salary. To this end, we propose a new feature screening method to select important skill words for interval-valued response. This method enjoys some merits. First, the marginal utility for feature screening is based on the difference of estimated distribution functions via nonparametric maximum likelihood estimation, which sufficiently use the interval information. Second, it is model-free and robust to outliers. Third, the sure screening property is also theoretically established. Numerical simulations show that the new method using the interval information is more efficient to select important predictors than the methods only based on the single point of the interval. In the real data analysis, we study the text data of job advertisements for data scientists and data analysts in a major Chinese online job posting website, and explore the important skill words for the salary. We find that the skill words like deep learning, recommendation algorithm, TensorFlow can boost the salary while the words like data collection, data summary, Excel may negatively contribute to the salary.