CEVM: Constrained Evidential Vocabulary Maintenance Policy for CBR Systems

Abstract

The maintenance of Case-Based Reasoning (CBR) systems has attracted increasing interest within current research since they proved high-quality results in different real-world domains. This kind of systems stores previous experiences, which are described by a vocabulary (e.g., attributes), incrementally in a case base. Actually, the vocabulary presents one among the most important maintenance targets, since it highly contributes in providing accurate solutions and in improving systems' performance, especially within high-dimensional domains. However, there is no policy, in the literature, that offers the ability to exploit prior knowledge (e.g., given by domain-experts) during the maintenance of features describing cases. In this paper, we propose a flexible policy for the most relevant attribute selection based on the attribute clustering concept. This new policy is able, on the one hand, to manage uncertainty using the belief function theory based tools, and on the other hand, to make use of domain-experts knowledge in form of pairwise constraints: If two attributes offer the same information without any added-value, then a Must-link constraint between them is generated. Otherwise, if there is no relation between them and they offer different information, then a Cannot-link constraint between them is created.

CIMMEP: constrained integrated method for CBR maintenance based on evidential policies

“…CIMMEP is characterized by an alternation of two main phases so as to provide a trade-off between accurate maintenance tasks for CB and Vocabulary knowledge containers, while exploiting prior knowledge available for both of them. The first phase, which is inspired from our preliminary work described in [33], concerns vocabulary maintenance under constraints, where the second phase which regards CBM under constraints uses steps of a new weighted version of our preliminary work presented in [19]. Therefore, we present, in what follows, our constrained vocabulary maintenance strategy (Section 4.1), our new weighted and constrained policy for CBM (Section 4.2), and finally our main proposal regarding the new constrained and integrated CIMMEP method for both CB and vocabulary maintenance (Section 4.3).…”

“…We note from this table that our CIMMEP method, which alternates between constrained learning and maintenance, is able to maintain CB and Vocabulary knowledge while exploiting prior available knowledge, not sensible to noisy data, and capable to manage uncertainty under the belief function framework. This comparative study has been made in front of the six following maintenance policies: CNN [9], ECTD [17], CECTD [19], ReliefF [37], EVM [20], and CEVM [33].…”

“…To generate Must-link and Cannot-link constraints, authors in [19] and [33] propose the two following modes:…”

“…where T hresh is a threshold that aims to answer to the question: "From which amount of membership certainty in [0, 1], we consider that the attributes/cases belong or not to the same cluster?". Since the alternate mode for constraints generation proved higher effectiveness in both [19] and [33], it will be used, during the experimental section, to execute and test our proposed methods.…”

Managing uncertainty during CBR systems vocabulary maintenance using Relational Evidential C-Means

“…Belief function theory or Evidence theory [25], [26] presents one of the most powerful tools for that matter. Some evidential clustering techniques such as k-evclus and EVCLUS [27], [28], have been used in [12] and [29] for features learning. However, other techniques [30] within this theory have been proposed and showed their effectiveness.…”