In software development, predicting the completion time of tasks is crucial for effective project management. However, due to the complexity of identifying process activities and thereby creating event logs for further analysis, traditional process prediction techniques cannot be readily applied to software data. A key challenge is that development activities are recorded as fine-grained file changes, spanning over multiple artifacts and connected to various other entities, such as the users who made the changes, the module membership, or other related links. In this paper, we present an approach to extract and analyze information from artifacts present in software repositories, allowing us to identify completion patterns exhibited in projects. Subsequently, we leverage time-series analysis to understand the evolution of these artifacts and predict their completion times based on file-level activity. We evaluate this approach against real-world data, showing its effectiveness and usefulness in predicting file completion times. Our work provides project managers with actionable insights into critical development areas, highlighting regions of unpredictability or potential delays, and improving decision-making.