Feature-based similarity estimation of construction subschedules

Application of reusable process templates can reduce the planning time and increase the efficiency of construction scheduling significantly. To support the template definition, recurring subschedules can be automatically identified and generalized as process patterns. For the recognition of similar subschedules it is appropriate to use inexact graph-based matching methods. However, these methods are generally very computationally intensive and therefore cannot be efficiently applied to big graphs in large databases. Suitable filtering techniques allow for the preselection of the candidate subschedules for the matching algorithms and thus for the reduction of the time complexity. Feature-based indexing is the most common filtering technique. A proper definition of features and a sufficient similarity approximation between the features and between the subschedules represent major challenges for a good indexing technique. This paper presents an overall concept for pattern recognition in construction schedules and focuses on the feature-based similarity estimation of subschedules.