Today, with the deepening of economic globalization and the further intensification of competition among enterprises, the role and functions of human resource managers of enterprises have undergone major changes. Their main responsibilities have shifted from administrative and logistical type work to becoming strategic partners in businesses. Therefore, competency-based human resource management is gradually attracting the attention of enterprises, which use the competency model to effectively identify the potential characteristics of individuals, thus dynamically realizing the consistency of human-work-organizational strategic goals. Based on the social reality of profound changes in human resource management in the era of big data, this study uses natural language processing and machine learning methods to dynamically mine 165,811 enterprise recruitment data for feature analysis and model construction of human resource management job capabilities. To capture the job profile data of job seekers, experts are required to observe and score the career profile according to the competency model, obtain training data, and then study the application of the competency model in resume screening, and build the competency matching model. The results show that among the variables extracted from resumes, whether they majored in human resources, internship experience, student officer experience, the size of the company they work for, their level and their highest level of education has a significant impact on overall ability characteristics, while working hours, reward experience, and education have no significant effect on ability characteristics. The accuracy, recall rate, accuracy, and Fl score of the ability prediction model built by machine learning are all above 70%, indicating that the model can effectively imitate the behavior of experts, evaluate the competency according to the resume, and realize the ability matching in the resume screening from small-scale manual operation to big data machine calculation.