Decision Tree Algorithm Based On Correlation Feature Weighting Choose Academic Relationship Classification Rule Extraction

The identification and classification of the relations between people is a research hotspot in recent years. Academic relationship research can pave the way for the large network of academic relations which is growing increasingly, and the classification of the academic relations is the most important for the academic relations research. The paper aims at extracting rules of classification of academic relations from huge applications by using decision trees algorithm, and proposes an improved decision tree algorithm which is based on correlation feature selection.In this paper, the academic relations between researchers are defined as relationship between teachers and students, together with co-author and co-project. According to the characteristics of the applications, we produce the structure of the applications and design feature extracting functions, then collect original input data for different relationship and get training dataset. Then we adopt C4.5algorithm to get decision trees and extract classification rules. Lastly, in order to improve the precision of decision trees, this paper proposes an improved C4.5algorithm, which is based on correlation feature selection method. In the proposed algorithm, some important features are selected at first, which are given a weight greater than the threshold so that these features can be priority selected when a decision tree is being produced. The result shows that the proposed C4.5algorithm can optimize the decision tree when the important features are not all included in the original trees. Meanwhile, the hybrid values of precision and recall rate of classification rules from the proposed C4.5algorithm is also improved.