Mathieu Wauters

A multivariate approach for top-down project control using earned value management

Project monitoring and the related decision to proceed to corrective action are crucial components of an integrated project management and control decision support system (DSS). Earned value management/earned schedule (EVM/ES) is a project control methodology that is typically applied for top-down project schedule control. However, traditional models do not correctly account for the multivariate nature of the EVM/ES measurement system. We therefore propose a multivariate model for EVM/ES, which implements a principal component analysis (PCA) on a simulated schedule control reference. During project progress, the real EVM/ES observations can then be projected onto these principal components. This allows for two new multivariate schedule control metrics (T2 and SPE) to be calculated, which can be dynamically monitored on project control charts. Using a computational experiment, we show that these multivariate schedule control metrics lead to performance improvements and practical advantages in comparison with traditional univariate EVM/ES models. Two multivariate project schedule control metrics are presented using well-known EVM metrics.Principal component analysis is used to build a correlation reference and to test the multivariate metrics.A large simulation study compares the novel control approach with the current best practice.Improvements are obtained in comparison with the traditional project control approach.

Study of the stability of earned value management forecasting

AbstractIn this paper, the authors focus on the stability of earned value management (EVM) forecasting methods. The contribution is threefold. First of all, a new criterion to measure stability that does not suffer from the disadvantages of the historically employed concept is proposed. Second, the stability of time and cost forecasting methods is compared and contrasted by means of a computational experiment on a topologically diverse data set. Throughout these experiments, the forecasting accuracy is reported as well, facilitating a trade-off between accuracy and stability. Finally, it is shown show that the novel stability metric can be used in practical environments using two real-life projects. The conclusions of this empirical validation are found to be largely in line with the computational results.

A Nearest Neighbour extension to project duration forecasting with Artificial Intelligence

In this paper, we provide a Nearest Neighbour based extension for project control forecasting with Earned Value Management. The k-Nearest Neighbour method is employed as a predictor and to reduce the size of a training set containing more similar observations. An Artificial Intelligence (AI) method then makes use of the reduced training set to predict the real duration of a project. Additionally, we report on the forecasting stability of the various AI methods and their hybrid Nearest Neighbour counterparts. A large computer experiment is set up to assess the forecasting accuracy and stability of the existing and newly proposed methods. The experiments indicate that the Nearest Neighbour technique yields the best stability results and is able to improve the AI methods when the training set is similar or not equal to the test set. Sensitivity checks vary the amount of historical data and number of neighbours, leading to the conclusion that having more historical data, from which the a relevant subset can be selected by means of the proposed Nearest Neighbour technique, is preferential.

A comparative study of Artificial Intelligence methods for project duration forecasting

We propose 5 AI methods for predicting a projects duration.A methodology including PCA, cross-validation and grid search is presented.A large computational experiment shows the good performance of the AI methods.A sensitivity analysis reveals the weakness of the proposed methods. This paper presents five Artificial Intelligence (AI) methods to predict the final duration of a project. A methodology that involves Monte Carlo simulation, Principal Component Analysis and cross-validation is proposed and can be applied by academics and practitioners. The performance of the AI methods is assessed by means of a large and topologically diverse dataset and is benchmarked against the best performing Earned Value Management/Earned Schedule (EVM/ES) methods. The results show that the AI methods outperform the EVM/ES methods if the training and test sets are at least similar to one another. Additionally, the AI methods report excellent early and mid-stage forecasting results. A robustness experiment gradually increases the discrepancy between the training and test sets and demonstrates the limitations of the newly proposed AI methods.

A study on complexity and uncertainty perception and solution strategies for the time/cost trade-off problem

In this article, the Discrete Time/Cost Tradeoff Problem (DTCTP) is revisited in light of a student experiment. Two solution strategies are distilled from the data of 444 participants and are structured by means of five building blocks: focus, activity criticality, ranking, intensity, and action. The impact of complexity and uncertainty on the cost objective is quantified in a large computational experiment. Specific attention is allocated to the influence of the actual and perceived complexity and uncertainty and the cost repercussions when reality and perception do not coincide.

Blended Learning in Project Management

This paper reports on results of experiments in the classroom with students following Project Management (PM) courses using a blended learning approach. It discusses the impact of communication on the student performance on business games as well as the advantage of the use of integrative case studies and their impact on the learning experience of these students. While the performance of students is obtained by marking their quantitative output on the business game or case exercise, their learning experience is measured through an analysis of the course evaluations filled out by these students. Diversity among the test population is guaranteed by testing our experiments on a sample of students with a different background, ranging from university students with or without a strong quantitative background but no practical experience, to MBA students at business schools and PM professionals participating in a PM training.