Abstract:This paper introduces a novel incremental preference elicitation-based approach to learning potentially non-monotonic preferences in multi-criteria sorting (MCS) problems, enabling decision makers to progressively provide assignment example preference information. Specifically, we first construct a max-margin optimization-based model to model potentially non-monotonic preferences and inconsistent assignment example preference information in each iteration of the incremental preference elicitation process. Using the optimal objective function value of the max-margin optimization-based model, we devise information amount measurement methods and question selection strategies to pinpoint the most informative alternative in each iteration within the framework of uncertainty sampling in active learning. Once the termination criterion is satisfied, the sorting result for non-reference alternatives can be determined through the use of two optimization models, i.e., the max-margin optimization-based model and the complexity controlling optimization model. Subsequently, two incremental preference elicitation-based algorithms are developed to learn potentially non-monotonic preferences, considering different termination criteria. Ultimately, we apply the proposed approach to a credit rating problem to elucidate the detailed implementation steps, and perform computational experiments on both artificial and real-world data sets to compare the proposed question selection strategies with several benchmark strategies.

What problem does this paper attempt to address?

This paper attempts to address two key challenges in the multi - criteria sorting (MCS) problem: **non - monotonic preferences** and **inconsistent assignment example preference information**. Specifically, the paper proposes an approach based on incremental preference elicitation to learn the non - monotonic preferences that decision - makers may exhibit in MCS problems and handle inconsistent assignment example preference information. ### Problem Background Multi - criteria sorting (MCS) is a rapidly developing branch in the field of multi - criteria decision - making. Its main objective is to assign a set of alternatives to several predefined ordered categories according to multiple criteria. MCS has wide applications in areas such as credit rating, inventory management, supplier evaluation, green building rating, and policy evaluation. ### Existing Problems 1. **Non - monotonic Preferences**: In some cases, decision - makers may show non - monotonic preferences for certain criteria, that is, as the value of the criterion increases, the preference does not necessarily increase or decrease monotonically. 2. **Inconsistent Assignment Example Preference Information**: During the incremental preference elicitation process, decision - makers may provide answers that are contradictory to the previously given preference information. ### Solutions To solve the above problems, the paper proposes a new method based on incremental preference elicitation, mainly including the following aspects: 1. **Maximum Margin Optimization Model**: A maximum margin optimization model is constructed to effectively handle potential non - monotonic preferences and inconsistent assignment example preference information. This model tolerates inconsistency by introducing auxiliary variables while maximizing the discrimination ability. \[ \max \alpha \varepsilon - (1 - \alpha)\frac{\sum_{a_i \in A_{R,t}} (\delta_i^+ + \delta_i^-)}{|A_{R,t}|} \] where \(\alpha \in (0,1)\) is used to balance the two objectives, \(\varepsilon\) is a parameter of discrimination ability, and \(\delta_i^+\) and \(\delta_i^-\) are auxiliary variables. 2. **Information Quantity Measurement Method and Problem Selection Strategy**: Using the optimal objective function value of the maximum margin optimization model, an information quantity measurement method and a problem selection strategy are designed to identify the most informative alternatives in each iteration. These strategies are based on the uncertainty sampling framework in active learning. 3. **Termination Criteria**: Two different termination criteria are proposed: - **Termination Criterion I**: Terminate when the number of questions answered by the decision - maker reaches a predetermined maximum. - **Termination Criterion II**: Terminate when the accuracy of the model reaches a predetermined target. 4. **Algorithm Implementation**: Two algorithms based on incremental preference elicitation are developed, considering different termination criteria respectively, to learn potential non - monotonic preferences. ### Application Examples The paper applies the proposed method to the credit rating problem, details the implementation steps, and verifies the effectiveness of the proposed problem selection strategy through computational experiments on artificial and real - world data sets. ### Summary The main contribution of this paper is to propose a new method based on incremental preference elicitation, which can effectively handle non - monotonic preferences and inconsistent assignment example preference information in MCS problems, thereby improving the efficiency and accuracy of the decision - making process.

An incremental preference elicitation-based approach to learning potentially non-monotonic preferences in multi-criteria sorting

Lexicographic optimization-based approaches to learning a representative model for multi-criteria sorting with non-monotonic criteria

Data-driven preference learning methods for value-driven multiple criteria sorting with interacting criteria

Data-driven Preference Learning Methods for Sorting Problems with Multiple Temporal Criteria

Deep preference learning for multiple criteria decision analysis

Learning MR-Sort Models from Non-Monotone Data

A nonlinear program model to obtain consensus priority vector in the analytic hierarchy process

Modeling Contingent Decision Behavior: A Bayesian Nonparametric Preference-Learning Approach

Selection of a representative sorting model in a preference disaggregation setting: a review of existing procedures, new proposals, and experimental comparison

Learning preference representations based on Choquet integrals for multicriteria decision making

Monotone-Value Neural Networks: Exploiting Preference Monotonicity in Combinatorial Assignment

Large-margin Feature Selection for Monotonic Classification

Hybrid Multi-Criteria Preference Ranking by Subsorting

Consensus-based TOPSIS-Sort-B for multi-criteria sorting in the context of group decision-making

Multistage Utility Preference Robust Optimization

Robust ordinal regression in preference learning and ranking

Data-driven preference learning in multiple criteria decision making in the evidential reasoning context

On Preference Learning Based on Sequential Bayesian Optimization with Pairwise Comparison

A Preference Elicitation Approach for the Ordered Weighted Averaging Criterion using Solution Choice Observations

Using Indifference Information in Robust Ordinal Regression

Interactive Evolutionary Multi-Objective Optimization via Learning-to-Rank