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.