International Journal on Advanced Science, Engineering and Information Technology, Vol. 11 (2021) No. 1, pages: 326-333, DOI:10.18517/ijaseit.11.1.6650

Using Variation in Weighting Criteria and String Size Matching on Hybrid Model Schema Matching

Edhy Sutanta, Erna Kumalasari Nurnawati, Rosalia Arum Kumalasanti

Abstract

Schema matching plays a vital role in the information integration process from heterogeneous databases. Generally, the process of schema matching is to receive input, which are two databases (one as the source and another as a target), to match similarity attributes, and generate output in the form of mapping the similarity of the attribute pairs that are declared suitable. Furthermore, the user will assess these attribute pairs to determine whether the results obtained are correct or still need to be revised. Our previous study developed a model and software prototype of hybrid schema matching using a combination of constraint-based method and instance-based method. In this study, the model improved by adding new features. This paper discusses the increasing effectiveness of adding the features to customize the weight of matching criteria and string sizes matching. The hybrid model's best effectiveness is obtained when the weight of instance is 0.286, the type is 0.238, width is 0.190, nullable is 0.143, unique is 0.095, and the domain is 0.048. The matching process using a bigger string size increases the model effectiveness with the highest precision of 97.66 when the string size interval is between (length-100) and (length+100). The best combination of weight and string size variation obtains 97.66% precision, a 99.90% recall, and an f-measure of 98.74%.

Keywords:

criteria; effectiveness; hybrid schema matching; string size variation; weight variation

Viewed: 1016 times (since Sept 4, 2017)

cite this paper     download