Abstract
M.Sc. (Computer Science)
Projects are undertaken in various parts of human life as a means for getting work
done and facilitating problem-solving (Kloppenborg, 2011). This is evident
throughout history with ancient examples of projects such as the construction of the
Great Pyramid of Giza in Egypt circa 2500 BC (Kozak-Holland, 2011). Modern
endeavours to collaborative productivity and problem-solving are also still driven by
projects. Much effort goes into planning projects, however, many of these plans are
unable to anticipate the execution needs of projects. As a result, many projects fail
despite having plans in place (Ewusi-Mensah, 2003). Therefore, there is an apparent
need for improving the aforementioned approach to planning projects, such that the
execution needs of projects are anticipated. In particular, there is a need for a
systematic process for forecasting project outcomes. This research study aims to
address the aforementioned need by proposing the Bolepi framework – a machine
learning framework (or methodology) for forecasting project outcomes. A hybrid
approach to research is employed in this research study, where the empirical and
implementation driven research approaches are combined in order to evaluate the
feasibility of Bolepi as a solution for the aforementioned problem. Furthermore,
Synergy (a prototype implementation of Bolepi) is presented in this research.
Synergy demonstrates how Bolepi works by forecasting the outcomes of construction
projects with the use of neuroevolution. Synergy was evaluated in order to determine
its feasibility as a tool for forecasting project outcomes. As Synergy’s initially
recorded performance did not meet the success requirements of this research,
bootstrap aggregating was employed as a performance enhancement method. This
enhanced Synergy’s performance to an acceptable level for this research, and as such
illustrated that Synergy is capable of forecasting project outcomes. As such, through
this research the Bolepi framework has been validated as a feasible framework (or
methodology) for use in forecasting project outcomes. Future work on this topic will
investigate potential areas of improvement for Bolepi, particularly in optimising the
architecture and design used for forecasting the outcomes of projects.