Abstract
The task of project management for creating an information system for monitoring changes in air ticket prices as a data-driven product with a machine learning component is considered. The relevance of the study is due to the high dynamics of airline pricing, the dispersion of price information among numerous sources, and the underdevelopment of the proactive monitoring and price forecasting segment in the Ukrainian and European markets, which makes independent price monitoring difficult for the average user. The developed solutions will allow Ukrainian consumers to prepare for the resumption of air transportation after the end of the war. The purpose of the study is to develop a model for managing such a project and a set of project management tools that take into account the specifics of data accumulation and the research nature of training predictive models. To achieve this goal, a systemic analysis of the subject area, a comparative analysis of IT project management methodologies, a SWOT analysis, the work breakdown structure method (WBS), and the logical framework approach (LFA) were used. The choice of a hybrid management methodology is justified, combining Scrum as the main development framework, CRISP-DM elements for the data and model lifecycle, Kanban for the support phase, as well as ISO 31000 and MSF for risk management. The classic three-phase Scrum structure is expanded to four phases (initialization, planning, development, support and development), reflecting the specifics of a data-driven product, with an emphasis on the early start of the data collection subsystem as a critical prerequisite for training predictive models. On this basis, a hierarchical WBS is built according to the product principle, which decomposes work into functional modules of the system down to the level of work packages. The SWOT analysis reveals the strengths and weaknesses of the project, opportunities and threats of the external environment, and the principles of effective management are substantiated. Using LFA, the subject area is formalized by building a problem tree. Prospects for further research are related to the development and validation of a predictive model of price dynamics on real data and the transition to more complex ML algorithms.