Full length articleAn improved visualization-based approach for project portfolio selection
Introduction
Multi-objective optimization approaches model the entities of a problem as well as preferences and requirements provided by decision makers in order to calculate one or more non-dominated optimized solutions that make the Pareto frontier for the problem. However, models may not capture entirely all user requirements and preferences, and so they produce optimum solutions in terms of the modeled problem only, instead of dealing with the real world problem. Interactive approaches may complement this scenario (Fisher, 1985). In the Project Portfolio Selection (PPS) problem, given a set of projects, an optimization software must determine when each project begins, in order to optimize technical criteria, such as reducing the portfolio makespan or determining minimum funding necessities. After that, some approaches use interactive visualization techniques to present these solutions to decision makers, in order to enable them to further explore the solution set.
In this work, we show how single-objective optimization problems may also benefit from interactive approaches. A single-objective optimization problem aims to construct a single best solution of a problem, i.e., it seeks the best value of a single objective function (Bandyopadhyay & Saha, 2013). Enabling decision makers to modify a solution helps them to adjust their preferences in a neighborhood of a proposed best solution. Besides, our visualization-based approach, based on interactive coordinated graphics, enables users to explore alternative solutions in an incremental, reversible and easy-to-learn way. Our main contributions are:
- 1.
An interactive visualization approach for insertion of decision maker's preferences into the optimization problem;
- 2.
An interactive coordinated view of costs, risks, man-hours and individual projects allocation in time, based on timelines, barcharts, and line charts.
- 3.
An alternative layout algorithm for timelines, which allows for dragging events (in this context, portfolio projects) by increasing the interaction area.
The next sections are organized as follows. First, we review interactive approaches and visualization solutions, followed by a more precise description of the PPS we have at hand. After that, we explain our visualization-based methodology for dealing with our single-objective PPS problem, which includes user, data, tasks, and insights characterizations, as well as the definition of interactive charts that provide such insights to users. Then we present the results of a heuristic evaluation of usability. Besides, we show the use of our approach in a real world scenario, and report on preliminary user opinions about it. Lastly, we conclude and present ideas for future work. It is worth noting that this paper is an extended version of another published work of our research team (Silva et al. 2016b).
Section snippets
Theoretical background
We first present a literature overview covering some approaches for human interfaces when interacting with PPS solutions. The second subsection gives more precise information about the PPS we applied the visualization techniques to.
Methodology
Inspired by Fisher's and Nowak's arguments in favor of adopting interactive techniques in conjunction with multicriteria decision making approaches (Nowak, 2013, Fisher, 1985), we propose the use of visualization-based interactive optimization methods for our single-objective PPS problem. In our proposal, two phases alternate until a decision maker is satisfied. In the first phase, the interactive one, we present graphical versions of the current portfolio and values of variables related to it,
Results and discussion
In this section we discuss how our approach helps users answer questions Q1 to Q4. We discuss these questions and their respective insights based on synthetic scenarios. Also, we report the results of a heuristic evaluation applied to our software, and describe preliminary client feedback about our prototype when operating on real data.
Conclusion
This paper presented a visualization-based approach for helping decision makers to solve project portfolio selection problems construed as single-criteria decision making problems. Our interactive timeline enables users to express their preferences about project starting times. At the same time, our coordinated cost and risk charts are instrumental to help them analyze the impact of these preferences on yearly costs, and consequently on portfolio feasibility. Our man-hour charts also enable
Acknowledgements
We thank ANEEL and AES for their financial support to our research.
References (15)
Project portfolio selection using interactive approach
Procedia Engineering
(2013)- et al.
Interactive multicriteria methods in portfolio decision analysis
- et al.
Unsupervised classification – Similarity measures, classical and metaheuristic approaches, and applications
(2013) Interactive optimization
Annals of Operations Research
(1985)- et al.
An experimental comparison of two interactive visualization methods for multicriteria portfolio selection
- et al.
Human-centered aspects
- et al.
Visualizing the pareto frontier