Narrow Results

In the belief function theory, several measures of uncertainty have been introduced. One of their possible use is unreliable source discounting before the fusion stage. Two different measures of uncertainty exist which are the intrinsic and extrinsic ones. The intrinsic measure makes it possible to assess the source's confusion whereas the extrinsic one measures the contradiction between sources. In this paper, we associate both measures in order to estimate the global reliability of a source. This method, named Generic Discounting Approach (GDA), is proposed in two different versions: Weighted GDA and Exponent GDA. Those reliability measures are integrated into a classifier. The method was tested, against to some pioneer approaches, on several UCI datasets as well as on an urban image classification problem and showed very encouraging results.

Decision making by considering multiple information sources could provide interesting results. For that reason, fusion formalisms were a major concern in the belief function community. In this context, the Belief function theory allows information fusion thanks to its combinations tools that it integrates. Nevertheless, belief function theory highlights a limit in the merging of contradictory (conflictual) sources. Many authors tackled this problem offering contributions in this field. Unfortunately, no proposed operator has distinguished by its adequacy regardless the type of handled sources. In this paper, we demonstrate the limits of some referenced works and we diagnostic the issues origin. We propose a conflict management approach based on an extra-information that guides the treatment. We also integrate a generic associative base borrowed from the data mining domain in order to apply the adequate conflict management.

In modeling Multi-Criteria Decision Making (MCDM) problem, we usually assume that the decision maker is able to elicitate his preferences with precision and without difficulty. However, in many situations, the expert is unable to provide his assessment with certainty or he is unwilling to quantify his preferences. To deal with such situations, a new MCDM model under uncertainty is introduced. In fact, we focus here on the problem of modeling expert opinions despite the presence of incompleteness and uncertainty in their preference assessments. Besides, our proposed solution suggests to model these preferences qualitatively rather than exact numbers. Therefore, we propose to incorporate belief preference relations into a MCDM method. The expert assessments are then formulated as a belief function problem since this theory is considered as a useful tool to model expert judgments.

This paper proposes a new assignment solution based on the Generalized Bayes’ Theorem (GBT) which aims to establish the best matching between two sets of uncertain data. In order to estimate the effectiveness of the proposition, it is compared to the best credal assignment solutions and the well known Global Nearest Neighbor (GNN) algorithm, through synthetic data and a literature example of multi-target tracking scenarios. Given the same input data, the proposed solution gives better assignment results, especially when sensor imprecision increases. However, the proposed solution stills actually computationally more complex than the GNN and the solution proposed by Denoeux et al.

Belief function theory is one of broadly used numerical formalisms for uncertainty processing. It was also studied in its qualitative form. In this paper we study relationship between this qualitative belief function theory and modal logic. We present a sound system of axioms and show its completeness with respect to sematics considered.

In this paper, the analytic hierarchy process (AHP) method is extended to an uncertain environment where the uncertainty is represented by belief functions as interpreted in the transferable belief model (TBM). Our proposed approach, called belief AHP, is developed to help the decision maker to determine what the best alternatives are, considering multiple conflicting criteria where both alternatives and criteria may be soiled with imperfection. The Belief AHP method aims at comparing subsets of criteria and groups of alternatives in order to reduce the pair-wise comparisons number. Furthermore, to handle uncertainty that may appear in the comparison procedure, we use basic belief assignments (BBA) instead of exact ratios to elicitate expert preferences. Finally, to illustrate the feasibility of our approach and to judge its performances, we have applied our proposed method on a real application problem: we have considered the polyvinyl chloride (PVC) life cycle assessment especially the end of life phase.

Learning Classifier Systems (LCSs) are a kind of evolutionary machine learning algorithms that provide highly adaptive components to deal with real world problems. They have been widely used in resolving complex problems such as decision making and classification. LCSs are flexible algorithms that are able to construct, incrementally, a set of rules and evolve them through the Evolutionary Algorithm (EA). Despite their efficiency, LCSs are not capable of handling imperfect information, which may lead to reduced performance in terms of classification accuracy. We propose a new accuracy-based Michigan-style LCS that integrates the belief function theory in the supervised classifier system. The belief function or evidence theory represents an efficient framework for treating imperfect information. The new approach shows promising results in real world classification problems.

Nowadays, machine learning is considered as one of the most popular fields in computer science that have shown a great success in the classification and prediction areas. In this paper, we focus on Artificial Immune Recognition System (AIRS) which is a supervised learning method inspired by the immune system metaphors. Such technique has been applied in various areas and has shown promising prediction results. Yet, different versions of AIRS methods have been proposed such as AIRS2 and AIRS3. Nevertheless, these two versions have some limits which are mainly related to their inability to work under uncertainty. This is considered as a big challenge in real-world classification problems. For this purpose, this paper contributes to the problem of covering uncertainty in the classification process using the belief function theory where a new machine learning approach called WE-AIRS (Weighted Evidential AIRS) is proposed. In WE-AIRS, the number of training antigens represented by each memory cell is taken into account and the classification of the antigens is performed based on their derived weights. The performance of the new weighted evidential method is validated on five real world data sets and compared to the other traditional AIRS versions.

In this paper, we deal with the problem of learning decision rules from partially uncertain data based on rough sets. The uncertainty exists in the decision attribute and not in condition attribute values of the decision system. This latter is represented by the belief function theory. So, we will adapt the basic concepts of rough sets in order to generate rules, denoted belief decision rules.

Maintaining the vocabulary of case knowledge within Case Based Reasoning (CBR) presents a crucial task to ensure a high-quality problem-solving and to improve retrieval performance for large-scale CBR systems. To do, we propose, in this paper, a method that manages uncertainty while selecting the best attributes characterizing case knowledge by using belief function theory. Actually, this method is based on a new evidential attribute clustering technique to eliminate redundant and noisy attributes describing cases.

One of the most promising approaches in the field of Recommender Systems (RSs) is Collaborative Filtering (CF). CF techniques are commonly divided in the two general classes of memory-based and model-based. A wise strategy would be to combine these two methods to increase their performance while leveling out the weakness of each one. Otherwise, the uncertainty pervaded throughout the different steps of the recommendation process should not be ignored. Handling uncertainty is very challenging and important for more reliable and intelligible predictions. That is why, we propose in this paper a new CF approach which combines these two categories under the belief function theory while dealing with the uncertainty pervaded in the prediction process. The effectiveness of our proposal is validated on a real-word data set and compared to state-of-the-art CF approaches under certain and uncertain frameworks.