Test Graph for SDG Inference Engine and Its Demonstration
WU Chong-guang · Acta Simulata Systematica Sinica · 2003
During the research and application of SDG (signed directed graph) in solving safety problems of large-scale complex system, its very important to find an efficient and credible way to assess inference Engines power. We propose a series of standard and unified test methods based on structured graphs, by which inference engines capacities, such as veracity, completeness, speed, and resolution, can be tested systematically. Through adjusting scale, status generation, relationship, position and shape of node and route, we create a flexible method to generate a set of structured graphs which are suitable for various testing application. Besides, software and hardware environmental requirements of engine test have been discussed, and some suggestion of its reasonable scale and test guideline are given. During application, we recommend that the test should be carried on with two scales, e.g. peak and average test, to evaluate its performances respectively under extreme status and normal status. At the end of this article, we give a demonstration of testing the SDG inference engine we developed, and the result proves that its practical. As for some problems of our engine which are found in test, some improvement measurements have been proposed.