A personalized counseling system using case-based reasoning with neural symbolic feature weighting (CANSY)

Abstract

In this article, we introduce a personalized counseling system based on context mining. As a technique for context mining, we have developed an algorithm called CANSY. It adopts trained neural networks for feature weighting and a value difference metric in order to measure distances between all possible values of symbolic features. CANSY plays a core role in classifying and presenting most similar cases from a case base. Experimental results show that CANSY along with a rule base can provide personalized information with a relatively high level of accuracy, and it is capable of recommending appropriate products or services.

How Case-Based Reasoning Explains Neural Networks: A Theoretical Analysis of XAI Using Post-Hoc Explanation-by-Example from a Survey of ANN-CBR Twin-Systems

“…Hence, it cannot be considered to be a twin-system; however, it does show the potential for local-weighting methods. Later work extends global weighting schemes to datasets having symbolic feature-values [53,54]. Overall, the global methods tend to produce very similar results for different values of k but the results from local methods are found to be demonstrably better.…”

“…ANN-CBR twins are a special-case of a hybrid system, that combines ANN and CBR modules, when both accuracy and interpretability are primary requirements of the overall system. Though ANNs and CBR were coupled as early hybrid systems [43][44][45], it is not really until the late-1990s that "true" twins emerge [20,[46][47][48][49][50][51][52][53][54]. Fig.…”

“…Around 1999, a South Korean group working at the Korean Advanced Institute of Science and Technology (KAIST), explored a range of feature-weighting techniques in comparative tests of the twin-system idea, in a framework they called "Memory Based Neural Reasoning" [46][47][48][49][50][51][52][53][54]. Shin and colleagues [46,47,51] paired MLPs with CBR operating over the same dataset, proposing that this hybrid system "can give examplebased explanations together with prediction values of the neural network" [47, p.637].…”

Loss and gain functions for CBR retrieval

“…In this example the weights were assigned by an expert in the field. The valuation of weights is a crucial part of this process, and various valuation approaches have been proposed [49,22,56,24].…”