An expert system is being developed to help assure the quality of data in the consumer price index (CPI). The system replicates the reasoning of commodity analysis, determining if a substitute product is comparable to a product priced in the previous month. If comparable, its price can be considered in producing the CPI. The system reasons with expertise directly entered by commodity analysts. To help the commodity analysts serve as their own knowledge engineers, they are provided with a skeletal system to which they add their knowledge of particular product categories. The system was tested on six months of CPI data. Results indicate that commodity analysts can accurately encode their own expertise and use the technology to detect their own oversights. A primary function of the project is to document and refine analyst expertise that is otherwise implicit. This makes the knowledge permanent and represents it in a way that allows it to be applied consistently.
展开▼
