Visualization of Similarity Models for CBR Comprehension and Maintenance

Modeling similarity measures in Case-Based Reasoning systems is a critical and multifaceted task. Typically, similarity measures are manually defined, often with the input of domain experts or utilizing machine learning methods. These measures are then subjected to evaluation processes that include metrics that assess the properties of the retrieval and reuse processes on the case base. In this paper, we present SimViz, an exploratory visualization tool aimed at understanding and identifying errors in both data and similarity measures. Our tool represents an instance of Explainable Case-Based Reasoning, enabling interactive visualization through heatmaps and histograms and assisting in case comparison. These visualizations provide insight into the similarity between local and global attributes across different case representations.