Generative-Graph Method for Project Duration Estimation

Services are often run as projects. Examples of services that are typically managed as projects include software development, construction, and marketing, among others. In this work, we introduce a novel method for the early estimation of project duration. We propose using historical data of tasks execution sequences in past projects to derive a probability graph that encapsulates the precedence structure in task execution. In contrast to conventional methods that estimate project duration simply on the basis of high level business requirements, this approach generates potential task graph ensuring more accurate estimation of project duration. We perform detailed sensitivity analysis to evaluate effectiveness of our approach using synthetic data. We also evaluate it on data from real software development projects.