A Correlation Evaluation Method for Complex Objects

Abstract

Data correlation evaluation is the basis for data analysis, and the academic community has proposed many indicators to evaluate it, such as the Euclidean distance, and angle cosine, and so on. However, it is difficult for these indicators to effectively express the correlation degree of complex objects. Using traffic intersections as an example, this article proposes an effective method to evaluate the correlation between complex objects. First, based on a large quantity of basic data, a standard data format describing traffic intersection attributes was proposed. Then, experienced engineers were asked to grade the correlation between intersections. Finally, the two intersection standard format datasets were used as model inputs, the engineer correlation rating as the output of the model, and the support vector machine model was employed for training. The results of this data experiment demonstrate that the trained model can effectively express the correlation degree between traffic intersections, and therefore proves the validity of the method.