Design and implementation of an expert system for data sanity checking

Abstract

The authors discuss the design and implementation of an expert system developed for automated data sanity checking and archiving. Traditionally data are collected and stored after some syntax checking. Then the offline data are checked for errors and omissions to 'clean' the database. The expert system discussed has been designed to perform these functions on an online basis and store only the 'clean' data. This is achieved by continuously monitoring incoming data and making decisions as to what actions should be taken whenever abnormal data are flagged. The authors present a procedural description of the system which would allow the same process to be applied to any other kind of data. In particular, three phases of the project are described in detail. First is the design or conception phase in which the inference engine, the representation of the rules, and the handling of the data are discussed. The second phase, implementation, deals in its major part with acquiring the knowledge and expressing it in terms of system rules. In the last phase, methods for testing and evaluating the overall system are presented together with their outcomes. >