SINGLE FIELD GRAMMAR-BASED FEATURE CREATION WITH DISCOVERY ITEMS

摘要

A large number of all record linkage research has been restricted to the area of customer and patient records and not part information. Tasked with the mission to relate records from two disjoint sets A and B containing aircraft part data in order to perform data mining analysis, it was discovered that no relevant feature values were present. In this research no specific linkage approach is given rather we introduce a combination of practical techniques which include a grammar-based approach along with the concept of identifying discovery items to assist the data analyst in constructing relevant features from a single field in a relation for a data processing step. Our approach creates features from one record field; in this case a part description field. We select the part description field, eliminate the other fields and utilize only the description field as it was determined that the other fields would yield little to no assistance when attempting a record linkage.