A mixed similarity measure based on rough sets theory (MSM-R) and some experimental results for data classification problem.

Abstract

Mixed Similarity Measure plays an important role in the distance-based or similarity-based knowledge discovery and data mining problems such as classification, clustering... This paper aims to present more detailed studies on the Mixed Similarity Measure, which has attribute weights determined automatically and based on Rough sets theory (called Mixed Similarity Measure based on Rough sets theory -(MSM-R). Moreover, the paper presents the experimental method and the experimental results for classification problem using MSM-R on some UCI datasets, comparing results with the results of classification using Goodall’s measurement. Two proposed classification methods are k-nearest neighbors and decision tree (using C4.5 software). The experiment results show the effectiveness and practical applicability of the MSM-R in the real-world data classification problems.