Metric Generalization of Classification Algorithms Based on Formal Concept Analysisстатья
Информация о цитировании статьи получена из
Scopus
Статья опубликована в журнале из списка Web of Science и/или Scopus
Дата последнего поиска статьи во внешних источниках: 4 мая 2020 г.
Аннотация:Classification algorithms based on formal concept analysis (FCA) differ in how they process nonbinary descriptions of objects: using them directly as given or after transforming them to binary attributes by a scaling procedure. A common weakness of classifiers of the second type is that they forget the metric structure of the initial attribute space. The main idea of this article is how to utilize the original metric information alongside with order-theoretical relations between objects and attributes. The metric approach substantially reduces the number of classification failures and provides additional information about the objects, thus opening new options for classifier construction. A classifier model is proposed that generalizes some of the existing FCA-based classification algorithm and opens new possibilities for their modification. The article also considers an alternative metric-based approach: specifically, introduction of a distance measure between formal concepts and its use for classifier modification.