A Gravity Matching Area Selection Method Based on GA-Bagging-SVM

To improve the gravity matching accuracy, it is very important to build up the direct relationship between distribution characteristics of gravity anomaly map and gravity matching accuracy. This paper proposed a gravity directional matching area selection method based on GA-Bagging-SVM. Firstly, the heading was divided into four main directions, and a gravity adaptability analysis dataset was established for each main direction. The feature vectors were composed of gravity characteristic parameters, and the sample labels were determined by random test-line simulation experiments. Then, a GA-Bagging-SVM ensemble classifier was designed to establish a mapping relationship between the gravity characteristic parameters and the gravity matching accuracy in different main directions and classify local areas into matching and non-matching areas. The simulation results show that the test set classification accuracy of GA-Bagging-SVM is larger than 90%. The average matching error of the selected matching areas is smaller than 0.7 grid. The proposed method can effectively select the matching areas for different headings.