Article Preview
Top1. Introduction
The case base reasoning (CBR) methodology uses past experiences to compare situations or problems and solves a new one. Its reasoner, remembers a preceding case similar to the present one and uses that to solve it. This methodology developed to deal with uncertainty, approximate reasoning and exploit knowledge domain.
It uses reasoning by analogy, which is a simple and practical technique that solves new problems by comparing them to ones that have already been solved in the past, to fit a new problem solution. It has been widely used in different domains as industry, health, banking, etc.., (Sharma & Sharma, 2020; El-Sappagh & Elmogy, 2020; Rohini, 2021; Shoaip et al., 2019). However, as major issues, we will cite (Elmousalami, 2020):
- •
Processing of a larger number of cases in memory, may affect the retrieval efficiency if the size of the case-base grows to a worrying level.
- •
The retrieval does not necessarily find a concrete solution for a problem. It usually proposes only a set of possible solutions that need adaptation.
- •
The retrieval of several similar cases involves looking for a strategy to choose the best one.
To overcome these issues, different techniques dealing with information retrieval, have been proposed providing alternative approaches (Sharma & Sharma, 2020; Mehli et al., 2021; Wang et al., 2019, Wotawa et al., 2019) which are well adapted for knowledge and reasoning, such as artificial neural networks (ANNs) (Sharma & Mehrotra, 2021), genetic algorithms (Bhattacharya & Machacha, 2021), fuzzy logic (Dong et al., 2021), nearest neighbor matching (Dong & Lu, 2019), ontologies (Kaoura et al., 2019), rule based reasoning (RBR) (Berka, 2020) and data mining methods such as classification and others (Mansoul & Atmani, 2017), (Banu et al., 2019), (Ayed et al., 2021; Gu et al., 2021; Louati et al., 2021). Moreover, other works were concentrated on integrating various techniques and conducted to “hybrid reasoning methodologies” or “hybrid systems”. Thus, the use of hybrid approaches has reached different domains and became an issue of current concern in CBR research (Bannour et al., 2021).
In this context, our paper provides an approach and an overall flowchart that guides CBR along with a review of the four-step cycle that characterizes the methodology (retrieve, reuse, revise and retrain), followed by a specific experimentation on data uploaded from University of California, Irvine, School of Information and Computer Science (UCI), Machine Learning Repository, to show the relevance of our study. The proposed approach operates a reducing of the case-base which will be proposed to CBR to extract the most suitable solution for a new problem. The reduced case-base (space search) demonstrates how this approach can be applied to support CBR retrieval and significantly lighten the adaptation task to avoid going through the whole case-base.
The remainder of this paper is organized as follows. Section 2 presents a briefly the main notions of our subject and the literature review. In section 3, a presentation of the methodology adopted and detailed description of our approach. In section 4, we present analysis and discussion of results collected from a comparative study with a real testing database in order to assess our approach. Finally, some conclusions are drawn and further associated aspects of our approach are presented in section 5.