Abstract
We consider imprecise evaluation of alternatives in multiple criteria ranking problems. The imprecise evaluations are represented by n-point intervals which are defined by the largest interval of possible evaluations and by its subintervals sequentially nested one in another. This sequence of subintervals is associated with an increasing sequence of plausibility, such that the plausibility of a subinterval is greater than the plausibility of the subinterval containing it. We explain the intuition that stands behind this proposal, and we show the advantage of n-point intervals compared to other methods dealing with imprecise evaluations. Although n-point intervals can be applied in any multiple criteria decision aiding (MCDA) method, in this paper, we focus on their application in robust ordinal regression which, unlike other MCDA methods, takes into account all compatible instances of an adopted preference model, which reproduce an indirect preference information provided by the decision maker. An illustrative example shows how the method can be applied in practice.
Similar content being viewed by others
Notes
Let us remember that by c we denote the number of nested intervals in the n-point intervals, which correspond to c levels of plausibility.
References
Corrente, S., Greco, S., & Słowiński, R. (2012). Multiple criteria hierarchy process in robust ordinal regression. Decision Support Systems, 53(3), 660–674.
Corrente, S., Greco, S., Kadziński, M., & Słowiński, R. (2013). Robust ordinal regression in preference learning and ranking. Machine Learning, 93, 381–422.
Czyżak, P., & Słowiński, R. (1997). A concordance-discordance approach to multi-criteria ranking of actions with fuzzy evaluations. In J. Climaco (Ed.), Multicriteria analysis (pp. 85–93). Berlin: Springer.
Dembczynski, K., Greco, S., & Słowiński, R. (2009). Rough set approach to multiple criteria classification with imprecise evaluations and assignments. European Journal of Operational Research, 198(2), 626–636.
Dong, Y., & Herrera-Viedma, E. (2015). Consistency-driven automatic methodology to set interval numerical scales of 2-tuple linguistic term sets and its use in the linguistic GDM with preference relation. Cybernetics, IEEE Transactions on, 45(4), 780–792.
Dubois, D. (2011). The role of fuzzy sets in decision sciences: Old techniques and new directions. Fuzzy Sets and Systems, 184(1), 3–28.
Figueira, J., Greco, S., & Ehrgott, M. (2005). Multiple criteria decision analysis: State of the art surveys. Berlin: Springer.
Fishburn, P. C. (1985). Interval orders and interval graphs. New York: Wiley.
Giarlotta, A., & Greco, S. (2013). Necessary and possible preference structures. Journal of Mathematical Economics, 49(2), 163–172.
Greco, S., Matarazzo, B., & Słowiński, R. (2001). Rough sets theory for multicriteria decision analysis. European Journal of Operational Research, 129(1), 1–47.
Greco, S., Mousseau, V., & Słowiński, R. (2008). Ordinal regression revisited: Multiple criteria ranking using a set of additive value functions. European Journal of Operational Research, 191(2), 416–436.
Herrera, F., Alonso, S., Chiclana, F., & Herrera-Viedma, E. (2009). Computing with words in decision making: Foundations, trends and prospects. Fuzzy Optimization and Decision Making, 8(4), 337–364.
Hurwicz, L. (1951). A class of criteria for decision making under ignorance. Cowles series, Statistics, 356.
Jacquet-Lagreze, E., & Siskos, J. (2001). Preference disaggregation: 20 years of MCDA experience. European Journal of Operational Research, 130(2), 233–245.
Keeney, R. L., & Raiffa, H. (1993). Decisions with multiple objectives: Preferences and value tradeoffs. New York: Wiley.
Lahdelma, R., Hokkanen, J., & Salminen, P. (1998). SMAA-stochastic multiobjective acceptability analysis. European Journal of Operational Research, 106(1), 137–143.
Moskowitz, H., Preckel, P. V., & Yang, A. (1993). Decision analysis with incomplete utility and probability information. Operations Research, 106(1), 137–143.
Ozturk, M., Pirlot, M., & Tsoukias, A. (2011). Representing preferences using intervals. Artificial Intelligence, 175(7–8), 1194–1222.
Park, K. S., & Kim, S. J. (1997). Tools for interactive multiattribute decision-making with incompletely identified information. European Journal of Operational Research, 98(1), 111–123.
Roy, B. (1993). Decision science or decision-aid science? European Journal of Operational Research, 66(2), 184–203.
Roy, B. (1996). Multicriteria methodology for decision aiding. Dordrecht: Kluwer.
Roy, B., & Słowiński, R. (2013). Questions guiding the choice of a multicriteria decision aiding method. EURO Journal on Decision Processes, 1(1), 69–97.
Słowiński, R., Greco, S., & Matarazzo, B. (2009). Rough sets in decision making. In R. A. Meyers (Ed.), Encyclopedia of complexity and systems science (pp. 7753–7786). New York: Springer.
Weber, M. (1987). Decision making with incomplete information. European Journal of Operational Research, 28(1), 44–57.
Zadeh, L. H. (1975). The concept of a linguistic variable and its application to approximate reasoning. Information Science, 1(8), 199–249.
Acknowledgments
The first and the second authors wish to acknowledge funding by the “FIR of the University of Catania BCAEA3 New developments in Multiple Criteria Decision Aiding (MCDA) and their application to territorial competitiveness”
Author information
Authors and Affiliations
Corresponding author
Appendix
Appendix
Proof of Proposition 4.1
-
1.
Let \(a\in A,\) and \(i,k\in \left\{ 1,\ldots ,n\right\} \) such that \(i\ge k\); this implies that \(g_{j}^{i}(a)\ge g_{j}^{k}(a),\; \forall j=1,\ldots ,m,\) and thus \(a\varDelta ^{(i,k)}a\). Therefore \(\varDelta ^{(i,k)}\) is reflexive. \(\square \)
-
2.
Let us suppose that \(a,b,c\in A\) such that \(a\varDelta ^{(i,k)}b\), \(b\varDelta ^{(i,k)}c\) and \(i\le k\) with \(i,k\in \left\{ 1,\ldots ,n\right\} \);
$$\begin{aligned} \left\{ \begin{array}{lll} a\varDelta ^{(i,k)}b &{} \Leftrightarrow &{} g_{j}^{i}(a)\ge g_{j}^{k}(b), \forall j\in J,\\ \\ b\varDelta ^{(i,k)}c &{} \Leftrightarrow &{} g_{j}^{i}(b)\ge g_{j}^{k}(c), \forall j\in J.\\ \end{array} \right. \end{aligned}$$Being \(i\le k\) we get that \(g_{j}^{k}(b)\ge g_{j}^{i}(b),\) and, consequently:
$$\begin{aligned}&g_{j}^{i}(a)\ge g_{j}^{k}(b)\ge g_{j}^{i}(b)\ge g_{j}^{k}(c), \forall j=1,\ldots ,m \Rightarrow g_{j}^{i}(a)\ge g_{j}^{k}(c),\\&\quad \forall j\in J \Leftrightarrow a\varDelta ^{(i,k)}c. \end{aligned}$$Thus \(\varDelta ^{(i,k)}\) is transitive.\(\square \)
-
3.
It follows from points 1 and 2 of this Proposition since a partial preorder is a reflexive and transitive binary relation. \(\square \)
-
4.
Let be \(a,b\in A\) and \(i,k,r,s\in \left\{ 1,\ldots ,n\right\} \) such that \(a\varDelta ^{(i,k)}b\), \(r\ge i\) and \(k\ge s\). Then
$$\begin{aligned} \left\{ \begin{array}{lll} a\varDelta ^{(i,k)}b &{} \Leftrightarrow &{} g_{j}^{i}(a)\ge g_{j}^{k}(b),\quad \forall j\in J,\\ \\ r\ge i &{} \Leftrightarrow &{} g_{j}^{r}(a)\ge g_{j}^{i}(a),\quad \forall j\in J, \\ \\ k\ge s &{} \Leftrightarrow &{} g_{j}^{k}(b)\ge g_{j}^{s}(b),\quad \forall j\in J. \end{array} \right. \end{aligned}$$From this it follows that:
$$\begin{aligned}&g_{j}^{r}(a)\ge g_{j}^{i}(a)\ge g_{j}^{k}(b)\ge g_{j}^{s}(b), \;\forall j\in J\Rightarrow g_{j}^{r}(a) \ge g_{j}^{s}(b),\\&\;\quad \forall j\in J\Leftrightarrow a{\varDelta }^{(r,s)}b.\;\; \end{aligned}$$\(\square \)
-
5.
Let \(a,b,c\in A\), and \(i,k,i_1,k_1\in \left\{ 1,\ldots ,n\right\} \) such that \(a\varDelta ^{(i,k)}b\), \(b\varDelta ^{(i_1,k_1)}c\) and \(k\ge i_1\). Then, we have:
$$\begin{aligned} \left\{ \begin{array}{lll} a\varDelta ^{(i,k)}b &{} \Leftrightarrow &{} g_{j}^{i}(a)\ge g_{j}^{k}(b),\quad \forall j\in J,\\ \\ b\varDelta ^{(i_1,k_1)}c &{} \Leftrightarrow &{} g_{j}^{i_1}(b)\ge g_{j}^{k_1}(c),\quad \forall j\in J,\\ \\ k\ge i_1 &{} \Leftrightarrow &{} g_{j}^{k}(b)\ge g_{j}^{i_1}(b), \quad \forall j\in J. \end{array} \right. \end{aligned}$$From this it follows that:
$$\begin{aligned} g_{j}^{i}(a)\ge g_{j}^{k}(b)\ge g_{j}^{i_1}(b)\ge g_{j}^{k_1}(c),\; \forall j\in J\Rightarrow g_{j}^{i}(a)\ge g_{j}^{k_1}(c),\;\forall j\in J \Leftrightarrow a\varDelta ^{(i,k_1)}c. \end{aligned}$$For point 4 of this Proposition, if \(r,s\in \left\{ 1,\ldots ,n\right\} \) such that \(r\ge i\) and \(s\le k_1\), then \(a\varDelta ^{(r,s)}c.\) \(\square \)
-
6.
We have said that \(\varDelta =\displaystyle \cap _{i=1}^{n}\varDelta ^{(i,i)}\); since \(\varDelta ^{(i,i)}\) is a partial preorder for point 3 of this Proposition, and since the intersection of partial preorders is a partial preoder, \(\varDelta \) is a partial preorder. \(\square \)
-
7.
Let \(a,b,c\in A\), and \(i,k\in \left\{ 1,\ldots ,n\right\} \), such that \(a\varDelta ^{(i,k)}b\), \(b\varDelta c\), \(s\ge i\) and \(k\ge t\). Then, we have:
$$\begin{aligned} \left\{ \begin{array}{lll} a\varDelta ^{(i,k)}b &{} \Leftrightarrow &{} g_{j}^{i}(a)\ge g_{j}^{k}(b),\quad \forall j\in J,\\ \\ b\varDelta c &{} \Leftrightarrow &{} g_{j}^{r}(b)\ge g_{j}^{r}(c), \quad \forall j\in J,\quad \forall r=1,\ldots ,n,\\ \\ s\ge i &{} \Leftrightarrow &{} g_{j}^{s}(a)\ge g_{j}^{i}(a), \quad \forall j\in J,\\ \\ k\ge t &{} \Leftrightarrow &{} g_{j}^{k}(c)\ge g_{j}^{t}(c),\quad \forall j\in J.\\ \end{array} \right. \end{aligned}$$From this it follows that:
$$\begin{aligned}&g_{j}^{s}(a)\ge g_{j}^{i}(a)\ge g_{j}^{k}(b)\ge g_{j}^{k}(c)\ge g_{j}^{t}(c),\; \forall j\in J\Rightarrow g_{j}^{s}(a) \ge g_{j}^{t}(c),\\&\quad \;\forall j\in J \Leftrightarrow a\varDelta ^{(s,t)}c. \; \end{aligned}$$\(\square \)
-
8.
Let \(a,b,c\in A\), and \(i,k\in \left\{ 1,\ldots ,n\right\} \), such that \(a\varDelta b\), \(b\varDelta ^{(i,k)} c\), \(s\ge i\) and \(k\ge t\). Then, we have:
$$\begin{aligned} \left\{ \begin{array}{lll} a\varDelta b &{} \Leftrightarrow &{} g_{j}^{r}(a)\ge g_{j}^{r}(b),\quad \forall j\in J,\quad \forall r=1,\ldots ,n,\\ \\ b\varDelta ^{(i,k)}c &{} \Leftrightarrow &{} g_{j}^{i}(b)\ge g_{j}^{k}(c), \quad \forall j\in J,\\ \\ s\ge i &{} \Leftrightarrow &{} g_{j}^{s}(a)\ge g_{j}^{i}(a),\quad \forall j\in J,\\ \\ k\ge t &{} \Leftrightarrow &{} g_{j}^{k}(c)\ge g_{j}^{t}(c),\quad \forall j\in J,\\ \end{array} \right. \end{aligned}$$From this it follows that:
$$\begin{aligned}&g_{j}^{s}(a)\ge g_{j}^{i}(a)\ge g_{j}^{i}(b)\ge g_{j}^{k}(c)\ge g_{j}^{t}(c),\;\quad \forall j\in J\Rightarrow g_{j}^{s}(a)\ge g_{j}^{t}(c),\\&\quad \;\forall j\in J \Leftrightarrow a\varDelta ^{(s,t)}c. \; \end{aligned}$$\(\square \)
Proof of Proposition 4.2
-
1.
For all \(i\in \left\{ 1,\ldots ,n\right\} \), from point 4 of Proposition 4.1, we have \(\varDelta ^{(1,n)}\subseteq \varDelta ^{(i,i)}\) because \(i\ge 1\) and \(i\le n\); from this follows that \(\varDelta ^{(1,n)}\;\subseteq \;\cap _{i=1}^{n}\varDelta ^{(i,i)}=\varDelta \), and this proves the first part of the Proposition.\(\square \)
For the same reason, \(\forall i\in \left\{ 1,\ldots ,n\right\} \) we have \(\varDelta ^{(i,i)}\;\subseteq \;\varDelta ^{(n,1)}\) because \(n\ge i\) and \(1\le i\). From this it follows that \(\varDelta =\cap _{i=1}^{n}\varDelta ^{(i,i)}\;\subseteq \;\varDelta ^{(n,1)}\), and this proves the second part of the Proposition.\(\square \)
-
2.
For all \(i,k\in \left\{ 1,\ldots ,n\right\} \), from point 4 of Proposition 4.1, since \(1\le i\) and \(k\le n\), we have \(\varDelta ^{(1,n)}\subseteq \varDelta ^{(i,k)}\) and \(\varDelta ^{(i,k)}\subseteq \varDelta ^{(n,1)}.\) In this way we obtain the thesis.\(\square \)
Proof of Proposition 5.1
-
1.
Let \(a\in A\) and \(i,k\in \left\{ 1,\ldots ,n\right\} \). From the definition, fictitious alternatives \(a^{(i)}\) and \(a^{(k)}\) are such that \(\forall j\in J\), and \(\forall r\in \left\{ 1,\ldots ,n\right\} ,\) \(g_{j}^{r}\left( a^{(i)}\right) =g_{j}^{i}(a)\), and \(g_{j}^{r}\left( a^{(k)}\right) \!=\!g_{j}^{k}(a)\). Since \(i\!\ge \!k\), \(\forall r\!\in \!\left\{ 1,\ldots ,n\right\} \) and \(\forall j\in J\), \(g_{j}^{r}(a^{(i)})\ge g_{j}^{r}(a^{(k)})\), and using the monotonicity of marginal value functions \(u_{j,r}(\cdot )\), we obtain \(\forall r\in \left\{ 1,\ldots ,n\right\} \), \(\forall j\in J\) \(u_{j}^{r}(g_{j}^{r}(a^{(i)}))\ge u_{j}^{r}(g_{j}^{r}(a^{(k)}))\); adding up with respect to j and r we obtain the thesis.\(\square \)
-
2.
We have seen that:
$$\begin{aligned} \displaystyle U(a^{(1)})= & {} \sum _{j=1}^{m}\left[ \sum _{i=1}^{n}u_{j}^{i}(g_{j}^{1}(a))\right] ,\;\;\;U(a)=\sum _{j=1}^{m}\left[ \sum _{i=1}^{n}u_{j}^{i}(g_{j}^{i}(a))\right] \\ \displaystyle U(a^{(n)})= & {} \sum _{j=1}^{m}\left[ \sum _{i=1}^{n}u_{j}^{i}(g_{j}^{n}(a))\right] . \end{aligned}$$\(\forall j\in J,\) and \(\forall i\in \left\{ 1,\ldots ,n\right\} \), since \(g_{j}^{1}(a)\le g_{j}^{i}(a)\le g_{j}^{n}(a)\) and by monotonicity of marginal value functions \(u_{j}^{i}(\cdot )\), we obtain:
$$\begin{aligned} u_{j}^{i}\left( g_{j}^{1}(a)\right) \le u_{j}^{i}\left( g_{j}^{i}(a)\right) \le u_{j}^{i}\left( g_{j}^{n}(a)\right) \end{aligned}$$and therefore adding up with respect to j and i we obtain the thesis, that is
$$\begin{aligned} U\left( a^{(1)}\right) \le U(a)\le U\left( a^{(n)}\right) . \end{aligned}$$\(\square \)
Proof of Proposition 6.1
-
1.
Let \(a,b\in A\), such that \(a\varDelta b\). This implies that \(g_{j}^{i}(a)\ge g_{j}^{i}(b), \forall j\in J, \forall i=1,\ldots ,n.\) We know that, for all \(U\in \mathcal{U}\):
$$\begin{aligned} U(a)=\sum _{j=1}^{m}\left[ \sum _{i=1}^{n}u_{j}^{i}(g_{j}^{i}(a))\right] . \end{aligned}$$From monotonicity of marginal value functions \(u_{j}^{i}(\cdot )\), we have that \(\forall j\in J,\) and \(\forall i=1,\ldots ,n,\) \(u_{j}^{i}(g_{j}^{i}(a))\ge u_{j}^{i}(g_{j}^{i}(b))\) and adding up with respect to indices j and i, we obtain \(U(a)\ge U(b)\) for all compatible value functions, thus we obtain the thesis.\(\square \)
-
2.
Let \(a,b\in A\), and \(i,k\in \left\{ 1,\ldots ,n\right\} ,\) such that \(a\varDelta ^{(i,k)}b\). This implies that \(g_{j}^{i}(a)\ge g_{j}^{k}(b), \forall j\in J.\) We know that, for all \(U\in \mathcal{U}\):
$$\begin{aligned} U(a^{(i)})=\sum _{j=1}^{m}\left[ \sum _{r=1}^{n}u_{j}^{r}(g_{j}^{i}(a))\right] ,\;\;\;U(b^{(k)})=\sum _{j=1}^{m}\left[ \sum _{r=1}^{n}u_{j}^{r}(g_{j}^{k}(b))\right] . \end{aligned}$$From the monotonicity of marginal value functions \(u_{j}^{r}(\cdot )\) we have that \(\forall j\in J,\) and \(\forall r=1,\ldots ,n,\) \(u_{j}^{r}(g_{j}^{i}(a))\ge u_{j}^{r}(g_{j}^{k}(b))\) and adding up with respect to indices j and r we obtain the thesis.\(\square \)
Proof of Proposition 6.2
-
1.
\(\forall a,b\in A\), \(\forall i,k\in \left\{ 1,\ldots ,n\right\} \), if \(a^{(i)}\) is at least as good as \(b^{(k)}\) for all compatible value functions \((a\succsim _{(i,k)}^{N}b)\), then there exists at least one compatible value function for which \(a^{(i)}\) is at least as good as \(b^{(k)}\) \((a\succsim _{(i,k)}^{P}b)\).\(\square \)
-
2.
It follows from point 1 of Proposition 5.1.\(\square \)
-
3.
Let \(a,b,c\in A\) and \(i,k\in \left\{ 1,\ldots ,n\right\} \), \(i\le k\), such that \(a\succsim _{(i,k)}^{N}b\) and \(b\succsim _{(i,k)}^{N}c\). This means that for all \(U\in \mathcal{U}\), \(U(a^{(i)})\ge U(b^{(k)})\) and \(U(b^{(i)})\ge U(c^{(k)})\). Since \(i\le k\), by point 1 of Proposition 5.1 we have that for all \(U\in \mathcal{U}\), \(U(a^{(i)})\ge U(b^{(k)})\ge U(b^{(i)})\ge U(c^{(k)})\) and, consequently, \(U(a^{(i)})\ge U(c^{(k)})\) for all \(U\in \mathcal{U}\), that is \(a\succsim _{(i,k)}^{N}c\).\(\square \)
-
4.
Let \(a,b\in A\), and \(i,k\in \left\{ 1,\ldots ,n\right\} \), such that \(a\not \succsim _{(i,k)}^{N}b.\) This means that \(\exists U\in \mathcal{U}: U(a^{(i)})<U(b^{(k)}).\) Therefore, \(b\succsim _{(k,i)}^{P}(a).\) \(\square \)
-
5.
Let \(a,b\in A\), and \(i,k\in \left\{ 1,\ldots ,n\right\} \) with \(i\ge k\) such that \(a\not \succsim _{(i,k)}^{P}b\). This means that for all \(U\in \mathcal{U}\), \(U(b^{(k)})>U(a^{(i)})\). Since \(i\ge k\), and from point 1 of Proposition 5.1, we obtain that for all \(U\in \mathcal{U},\) \(U(b^{(i)})\ge U(b^{(k)})>U(a^{(i)})\ge U(a^{(k)})\); thus for all \(U\in \mathcal{U}\), \(U(b^{(i)})> U(a^{(k)})\), therefore \(b\succsim _{(i,k)}^{N}a\) implying \(b\succsim _{(i,k)}^{P}a\) by point 1 of this Proposition. In this way, \(\succsim _{(i,k)}^{P}\) is strongly complete.
Let \(a,b,c\in A\), \(i,k\in \left\{ 1,\ldots ,n\right\} \), such that \(i\ge k\), \(a\not \succsim _{(i,k)}^{P}b\) and \(b\not \succsim _{(i,k)}^{P}c.\) Then, we have:
$$\begin{aligned} \left\{ \begin{array}{lll} a\not \succsim _{(i,k)}^{P}b &{} \Leftrightarrow &{} U(a^{(i)})<U(b^{(k)}),\quad \forall U\in \mathcal{U},\\ \\ b\not \succsim _{(i,k)}^{P}c &{} \Leftrightarrow &{} U(b^{(i)})<U(c^{(k)}), \quad \forall U\in \mathcal{U}.\\ \end{array} \right. \end{aligned}$$From this and from point 1 of Proposition 5.1 it follows that
$$\begin{aligned}&U(a^{(i)})<U(b^{(k)})\le U(b^{(i)})<U(c^{(k)}),\;\quad \forall U\in \mathcal{U}\Rightarrow U(a^{(i)}) < U(c^{(k)}), \\&\quad \;\forall U\in \mathcal{U}\Leftrightarrow a\not \succsim _{(i,k)}^{P}c. \end{aligned}$$This proves that \(\succsim _{(i,k)}^{P}\) is negatively transitive.\(\square \)
Proof of Proposition 6.3
-
1.
Let \(a,b\in A\), and \(i,k,i_1,k_1\in \left\{ 1,\ldots ,n\right\} \), such that \(i_1\ge i\), \(k_1\le k\) and \(a\succsim _{(i,k)}^{N}b\). Then, from point 1 of Proposition 5.1, we have:
$$\begin{aligned} \left\{ \begin{array}{lll} a\succsim _{(i,k)}^{N}b &{} \Leftrightarrow &{} U(a^{(i)})\ge U(b^{(k)}),\quad \forall U\in \mathcal{U},\\ \\ i_1\ge i &{} \Rightarrow &{} U(a^{(i_1)})\ge U(a^{(i)}),\quad \forall U\in \mathcal{U},\\ \\ k_1\le k &{} \Rightarrow &{} U(b^{(k_1)})\le U(b^{(k)}),\quad \forall U\in \mathcal{U}.\\ \end{array} \right. \end{aligned}$$Thus:
$$\begin{aligned}&U(a^{(i_1)})\ge U(a^{(i)})\ge U(b^{(k)})\ge U(b^{(k_1)}), \forall U\in \mathcal{U}\Rightarrow U(a^{(i_1)})\ge U(b^{(k_1)}), \\&\forall U\in \mathcal{U}\Leftrightarrow a\succsim _{(i_1,k_1)}^{N}b. \; \end{aligned}$$\(\square \)
-
2.
Let \(a,b\in A\), and \(i,k,i_1,k_1\in \left\{ 1,\ldots ,n\right\} \) such that \(i_1\ge i\), \(k_1\le k\) and \(a\succsim _{(i,k)}^{P}b\). Then, from point 1 of Proposition 5.1, we have:
$$\begin{aligned} \left\{ \begin{array}{l} a\succsim _{(i,k)}^{P}b \Leftrightarrow \exists U\in \mathcal{U}: U(a^{(i)})\ge U(b^{(k)}),\\ \\ i_1\ge i \Rightarrow U(a^{(i_1)})\ge U(a^{(i)}),\quad \forall U\in \mathcal{U},\\ \\ k_1\le k \Rightarrow U(b^{(k_1)})\le U(b^{(k)}),\quad \forall U\in \mathcal{U}.\\ \end{array} \right. \end{aligned}$$Thus:
$$\begin{aligned}&\exists U\in \mathcal{U}: U(a^{(i_1)})\ge U(a^{(i)})\ge U(b^{(k)})\ge U(b^{(k_1)})\Rightarrow \exists U\in \mathcal{U}: U(a^{(i_1)})\\&\quad \ge U(b^{(k_1)})\Leftrightarrow a\succsim _{(i_1,k_1)}^{P}b. \; \end{aligned}$$\(\square \)
-
3.
Let \(a,b\in A\) such that \(a\succsim _{(1,n)}^{N}b.\) This means that \(U(a^{(1)})\ge U(b^{(n)}),\) \(\forall U\in \mathcal{U}\). From point 2 of Proposition 5.1, we have that \(U(a)\ge U(a^{(1)})\ge U(b^{(n)})\ge U(b),\) \(\forall U\in \mathcal{U}\), and thus we obtain \(U(a)\ge U(b)\), \(\forall U\in \mathcal{U},\) that is \(a\succsim ^{N}b.\) In this way we proved that \(\succsim _{(1,n)}^{N}\;\subseteq \;\succsim ^{N}\).
Analogously, \(a\succsim ^{N}b\) means that \(U(a)\ge U(b),\) \(\forall U\in \mathcal{U}\); from point 2 of Proposition 5.1 we obtain \(U(a^{(n)})\ge U(a)\ge U(b)\ge U(b^{(1)})\), \(\forall U\in \mathcal{U}\), and thus we have \(U(a^{(n)})\ge U(b^{(1)})\), \(\forall U\in \mathcal{U},\) that is \(a\succsim _{(n,1)}^{N}b.\) In this way we proved that \(\succsim ^{N}\;\subseteq \;\succsim _{(n,1)}^{N}.\) \(\square \)
-
4.
Let \(a,b\in A\) such that \(a\succsim _{(1,n)}^{P}b.\) This means that \(\exists U\in \mathcal{U}: U(a^{(1)})\ge U(b^{(n)})\). From point 2 of Proposition 5.1 we have:
$$\begin{aligned} U(a)\ge U(a^{(1)})\ge U(b^{(n)})\ge U(b)\Rightarrow U(a)\ge U(b) \Leftrightarrow a\succsim ^{P}b. \end{aligned}$$In this way we proved that \(\succsim _{(1,n)}^{P}\;\subseteq \;\succsim ^{P}\).
Analogously, \(a\succsim ^{P}b\) means that \(\exists U\in \mathcal{U}: U(a)\ge U(b)\); from point 2 of Proposition 5.1 we obtain:
$$\begin{aligned} U(a^{(n)})\ge U(a)\ge U(b)\ge U(b^{(1)})\Rightarrow U(a^{(n)})\ge U(b^{(1)})\Leftrightarrow a\succsim _{(n,1)}^{P}b. \end{aligned}$$In this way we proved that \(\succsim ^{P}\;\subseteq \;\succsim _{(n,1)}^{P}.\) \(\square \)
Proposition 8.1 provides some further results regarding the imprecise necessary and possible preference relations.
Proposition 8.1
-
1.
If \(a\succsim _{(i,k)}^{N}b,\) \(b\succsim _{(i_1,k_1)}^{N}c,\) and \(k\ge i_1\), with \(i,k,i_1,k_1\in \left\{ 1,\ldots ,n\right\} \), then \(a\succsim _{(r,s)}^{N}c\), for all \(r,s\in \left\{ 1,\ldots ,n\right\} \) such that \(r\ge i\) and \(s\le k_1,\)
-
2.
If \(a\succsim _{(i,k)}^{N}b,\) \(b\succsim _{(i_1,k_1)}^{P}c,\) and \(k\ge i_1\), with \(i,k,i_1,k_1\in \left\{ 1,\ldots ,n\right\} \), then \(a\succsim _{(r,s)}^{P}c\), for all \(r,s\in \left\{ 1,\ldots ,n\right\} \) such that \(r\ge i\) and \(s\le k_1,\)
-
3.
If \(a\succsim _{(i,k)}^{P}b,\) \(b\succsim _{(i_1,k_1)}^{N}c,\) and \(k\ge i_1\), with \(i,k,i_1,k_1\in \left\{ 1,\ldots ,n\right\} \), then \(a\succsim _{(r,s)}^{P}c\), for all \(r,s\in \left\{ 1,\ldots ,n\right\} \) such that \(r\ge i\) and \(s\le k_1,\)
\(\square \)
Proof
-
1.
Let \(a,b,c\in A\) and \(i,k,i_1,k_1\in \left\{ 1,\ldots ,n\right\} \), such that \(a\succsim _{(i,k)}^{N}b\), \(b\succsim _{(i_1,k_1)}^{N}c\) and \(k\ge i_1\). Then we have:
$$\begin{aligned} \left\{ \begin{array}{l} a\succsim _{(i,k)}^{N}b \Leftrightarrow U(a^{(i)})\ge U(b^{(k)}),\quad \forall U\in \mathcal{U},\\ \\ b\succsim _{(i_1,k_1)}^{N}c \Leftrightarrow U(b^{(i_1)})\ge U(c^{(k_1)}),\quad \forall U\in \mathcal{U},\\ \\ k\ge i_1 \Rightarrow U(b^{(k)})\ge U(b^{(i_1)}) \end{array} \right. \end{aligned}$$From this it follows:
$$\begin{aligned}&U(a^{(i)})\ge U(b^{(k)})\ge U(b^{(i_1)})\ge U(c^{(k_1)}),\forall U\in \mathcal{U}\Rightarrow U(a^{(i)})\ge U(c^{(k_1)}), \\&\quad \forall U\in \mathcal{U}\Leftrightarrow a\succsim _{(i,k_1)}^{N}c. \end{aligned}$$Since \(a\succsim _{(i,k_1)}^{N}c\) and \(r,s\in \left\{ 1,\ldots ,n\right\} \): \(r\ge i\) and \(s\le k_1\), from point 1 of Proposition 6.3 we obtain \(a\succsim _{(r,s)}^{N}c.\) \(\square \)
-
2.
Let \(a,b,c\in A,\) and \(i,k,i_1,k_1\in \left\{ 1,\ldots ,n\right\} \), such that \(a\succsim _{(i,k)}^{N}b,\) \(b\succsim _{(i_1,k_1)}^{P}c,\) and \(k\ge i_1.\) Then we have:
$$\begin{aligned} \left\{ \begin{array}{l} a\succsim _{(i,k)}^{N}b \Leftrightarrow U(a^{(i)})\ge U(b^{(k)}),\quad \forall U\in \mathcal{U},\\ \\ b\succsim _{(i_1,k_1)}^{P}c \Leftrightarrow \exists U\in \mathcal{U}: U(b^{(i_1)})\ge U(c^{(k_1)}),\\ \\ k\ge i_1 \Rightarrow U(b^{(k)})\ge U(b^{(i_1)}),\quad \forall U\in \mathcal{U}. \end{array} \right. \end{aligned}$$It follows that:
$$\begin{aligned}&\exists U\in \mathcal{U}: U(a^{(i)})\ge U(b^{(k)})\ge U(b^{(i_1)})\ge U(c^{(k_1)})\Rightarrow \exists U\in \mathcal{U}: U(a^{(i)})\\&\quad \ge U(c^{(k_1)})\Leftrightarrow a\succsim _{(i,k_1)}^{P}c. \end{aligned}$$Since \(r,s\in \left\{ 1,\ldots ,n\right\} \): \(r\ge i\) and \(s\le k_1\), from point 2 of Proposition 6.3 we obtain \(a\succsim _{(r,s)}^{P}c.\) \(\square \)
-
3.
Let \(a,b,c\in A,\) \(i,k,i_1,k_1\in \left\{ 1,\ldots ,n\right\} \) such that \(a\succsim _{(i,k)}^{P}b,\) \(b\succsim _{(i_1,k_1)}^{N}c,\) and \(k\ge i_1.\) We have that:
$$\begin{aligned} \left\{ \begin{array}{l} a\succsim _{(i,k)}^{P}b \Leftrightarrow \exists U\in \mathcal{U}: U(a^{(i)})\ge U(b^{(k)}),\\ \\ b\succsim _{(i_1,k_1)}^{N}c \Leftrightarrow U(b^{(i_1)})\ge U(c^{(k_1)}), \quad \forall U\in \mathcal{U},\\ \\ k\ge i_1 \Rightarrow U(b^{(k)})\ge U(b^{(i_1)}), \quad \forall U\in \mathcal{U}. \end{array} \right. \end{aligned}$$From this it follows:
$$\begin{aligned}&\exists U\in \mathcal{U}: U(a^{(i)})\ge U(b^{(k)})\ge U(b^{(i_1)})\ge U(c^{(k_1)})\Rightarrow \exists U\in \mathcal{U}: U(a^{(i)}) \\&\quad \ge U(c^{(k_1)})\Leftrightarrow a\succsim _{(i,k_1)}^{P}c. \end{aligned}$$Since \(r,s\in \left\{ 1,\ldots ,n\right\} \) such that \(r\ge i\) and \(s\le k_1\), from point 2 of Proposition 6.3 we obtain \(a\succsim _{(r,s)}^{P}c.\) \(\square \)
Proposition 8.2 describes some properties involving the imprecise necessary and possible preference relations together with the classical necessary and possible preference relations.
Proposition 8.2
-
1.
Given \(a,b,c\in A\), \(i\in \left\{ 1,\ldots ,n\right\} \) such that \(a\succsim _{(i,n)}^{N} b\) and \(b\succsim ^{N}c\), then \(a\succsim _{(r,1)}^{N}c\), for all \(r\in \left\{ 1,\ldots ,n\right\} \) such that \(r\ge i\),
-
2.
Given \(a,b,c\in A\), \(k\in \left\{ 1,\ldots ,n\right\} \) such that \(a\succsim ^{N} b\) and \(b\succsim _{(1,k)}^{N}c\), then \(a\succsim _{(n,r)}^{N}c\), for all \(r\in \left\{ 1,\ldots ,n\right\} \) such that \(r\le k\),
-
3.
Given \(a,b,c\in A\), \(i\in \left\{ 1,\ldots ,n\right\} \) such that \(a\succsim _{(i,n)}^{P} b\) and \(b\succsim ^{N}c\), then \(a\succsim _{(r,1)}^{P}c\), for all \(r\in \left\{ 1,\ldots ,n\right\} \) such that \(r\ge i\),
-
4.
Given \(a,b,c\in A\), \(k\in \left\{ 1,\ldots ,n\right\} \) such that \(a\succsim ^{N} b\) and \(b\succsim _{(1,k)}^{P}c\), then \(a\succsim _{(n,r)}^{P}c\), for all \(r\in \left\{ 1,\ldots ,n\right\} \) such that \(r\le k\),
-
5.
Given \(a,b,c\in A\), \(i\in \left\{ 1,\ldots ,n\right\} \) such that \(a\succsim _{(i,n)}^{N} b\) and \(b\succsim ^{P}c\), then \(a\succsim _{(r,1)}^{P}c\), for all \(r\in \left\{ 1,\ldots ,n\right\} \) such that \(r\ge i\),
-
6.
Given \(a,b,c\in A\), \(k\in \left\{ 1,\ldots ,n\right\} \) such that \(a\succsim ^{P} b\) and \(b\succsim _{(1,k)}^{N}c\), then \(a\succsim _{(n,r)}^{P}c\), for all \(r\in \left\{ 1,\ldots ,n\right\} \) such that \(r\le k\).
Proof
-
1.
Let \(a,b,c\in A\) and \(i,r\in \left\{ 1,\ldots ,n\right\} \) such that \(a\succsim ^{N}_{(i,n)}b\), \(b\succsim ^{N}c\) and \(r\ge i\). Then we have:
$$\begin{aligned} \left\{ \begin{array}{ll} a\succsim ^{N}_{(i,n)}b \Leftrightarrow U(a^{(i)})\ge U(b^{(n)}), &{}\quad \forall U\in \mathcal{U},\\ \\ b\succsim ^{N}c \Leftrightarrow U(b)\ge U(c), &{}\quad \forall U\in \mathcal{U},\\ \\ r\ge i \Rightarrow U(a^{(r)})\ge U(a^{(i)}), &{}\quad \forall U\in \mathcal{U}. \end{array} \right. \end{aligned}$$It follows that, for all \(U\in \mathcal{U}\), \(U(a^{(r)})\ge U(a^{(i)})\ge U(b^{(n)})\ge U(b)\ge U(c)\ge U(c^{(1)})\) where \(U(b^{(n)})\ge U(b)\) and \(U(c)\ge U(c^{(1)})\) hold by point 2 of Proposition 5.1. Thus, for all \(U\in \mathcal{U}\) we obtain \(U(a^{(r)})\ge U(c^{(1)})\), and therefore \(a\succsim _{(r,1)}^{N}c.\) \(\square \) Points 2-6 can be proved analogously.\(\square \)
Rights and permissions
About this article
Cite this article
Corrente, S., Greco, S. & Słowiński, R. Handling imprecise evaluations in multiple criteria decision aiding and robust ordinal regression by n-point intervals. Fuzzy Optim Decis Making 16, 127–157 (2017). https://doi.org/10.1007/s10700-016-9244-x
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10700-016-9244-x