Using Qualitative Hypotheses to Identify Inaccurate Data Using Qualitative Hypotheses to Identify Inaccurate Data

Identifying inaccurate data has long been regarded as a di cult and signi cant problem in AI. In this paper, we present a new method to identify inaccurate data by considering qualitative connections among data. First, we put forward a new concept called support coe cient function(SCF). SCF can be used to extract, represent, and calculate qualitative e ects among data to enhance or depress the hypotheses that the inaccurate data are the expected ones, so that both fuzzy regions and membership degrees of inaccurate data can be dynamically determined on the basis of qualitative e ects among relating data, rather than on the basis of assumptions or prior probabilities supplied by human experts in advance. We have developed a practical system on infrared spectrum identi cation by using the method, and have run the system by hundreds of real spectra. The experimental results with the system are quite encouraging.