Visualized Performance Evaluation of 3D CAD Software Based on Rich Test Data and Radial Clustering Considering Credibility

The performance of 3D CAD software can be evaluated based on the test data obtained by the specifically developed automatic testing tool and testing cases. The values of the test data corresponding to the same performance index usually fluctuate due to the fluctuation of computer performance when a large number of repeated tests are implemented. To provide intuitive and convincing performance evaluation results, this paper proposes a visualized performance evaluation method for 3D CAD software based on rich test data and radial clustering considering credibility. Firstly, the credible scores of performance indexes are calculated based on the bounded probabilistic characterization of the rich test data collected by massive automatic test. Secondly, the multi-dimensional credible scores of the performance indexes of 3D CAD software are mapped to the elastic coefficients of the corresponding number of springs, and then radially projected onto a plane based on the force balance of all springs so that a Radviz radar map considering credibility is generated. Finally, the convincing performance evaluation results of 3D CAD software can be obtained through the clustering analysis of the Radviz radar map. A case study demonstrates the validity of the proposed method.