In most IT projects, business analysts do not use templates defined in industry standards but rather rework them to suit the needs of a specific project, creating a requirements architecture. The process of creating a requirements architecture must consider the specifics of the project and the set of techniques used to analyze the requirements. Therefore, the question of how to apply modern machine learning methods to generate recommendations based on the experience of practitioners is an important issue that needs to be addressed. The study uses the Association Rule Mining method, which has shown effectiveness in solving similar applied problems, to identify the implication relationship between the project context, the business analyst's profile, requirements analysis techniques, and requirements document structure. Based on the analysis of data collected during a survey of practicing IT specialists, significant patterns were identified, based on which recommendations for creating requirements analytical documents in IT projects were formed. The proposed approach is flexible and scalable, which allows its reuse for new data sets regarding the practices of business analysts in IT projects.