Write-offs in project management

The scope of this thesis is to research whether projects are a success or failure in terms of write-offs by using random forest as an algorithm for machine learning. Statistical analysis is first used for explaining the underlying variables for project success. The results are as follows: when the proportion of management hours to total hours involved in a project increases by 10%, the relative amount in write-off will increase by 10.3%. When one more managers are involved in the first stadium of a project, the relative amount in write-off will increase by 6.4%, whereas that 1 more manager is involved in the last stadium of a project, the relative amount in write-off shall decrease by 18.3%. When there is 1 more team member involved in a project, the relative amount in write-off will increase by 1.6%. When there is an increase of 10 euro in the average target rate, the relative amount in write-off will decrease by 2%, When there is 10% increase in the in team consistency, the relative amount in write-off will decrease by 5.12%, When there is a 1% increase in the total hours of a project, the relative amount in write-off will decrease by 33.4%. The relative amount in write-off will decrease by 8.72%, when the proportion of management hours to total hours involved in a project increases by 10% in the tax services than other line of services. Afterwards, random forest is used, which has an accuracy of 69% and the most important variables are total hours, total budget and total job days.