Multi-modal recommender system for predicting project manager performance within a competency-based framework.

Area of Science:

• Computational Social Science
• Human Resources Technology
• Performance Management Systems

Background:

• Traditional competency assessment for project managers is manual, time-consuming, and prone to evaluator bias.
• Accurate performance evaluation is crucial for professional development and organizational success.
• Existing methods lack efficiency and consistency in scoring diverse competencies.

Purpose of the Study:

• To develop and evaluate an automated approach for predicting project manager competency scores.
• To compare the effectiveness of various recommender system techniques in this context.
• To address the challenges of cold-starting in competency assessment.

Main Methods:

• Data fusion of demographic, profile, and historical assessment data.
• Natural Language Processing (NLP) for text data pre-processing.
• Comparison of content-based, demographic, collaborative, and hybrid filtering recommender systems.

Main Results:

• Content-based filtering achieved 81.03% precision for general competency scoring.
• Demographic filtering showed 54.05% precision for new project managers (cold-start).
• Content-based filtering demonstrated 85.79% precision for new competencies (cold-start).

Conclusions:

• Recommender systems offer a viable and efficient solution for automated competency assessment.
• Content-based filtering is particularly effective for both general and cold-start competency predictions.
• This approach can significantly streamline performance evaluation for project managers.