Abstract: In the domain of railway engineering construction management, the role of construction unit management is pivotal for achieving refined project management and promoting high-quality infrastructure development. Traditional approaches to screening participating construction units and personnel predominantly depend on single-source data and expert subjective judgment, which often lack scientific rigor, objectivity, and standardization in their criteria. This paper introduces a credit and service evaluation method based on multi-dimensional data analysis. Specifically, it utilizes source data such as compliance evaluations, performance assessments, deductions for adverse events, attendance records, safety and quality issues, and engineering suspension orders to construct a technical framework encompassing data collection, computation, access, and application. Additionally, the paper organizes existing data from the railway engineering construction management platform to create model training datasets while innovating data extraction techniques. Finally, a random forest algorithm model is trained to predict the creditworthiness and service quality of participating construction units or personnel. This approach leverages the value of extensive historical data, integrates information technology with business processes, and provides robust scientific support for the selection of railway engineering participants.