Study on Pilot Personality Selection with an SVM-Based Classifier

Purpose This paper intends to explore the feasibility of using statistical learning methods for learning and analyzing the data obtained from physiological tests and to offer novel ideas for the pilot selection and evaluation by investigating the personality traits of aviation professionals based on the results of the aforesaid exploration. Method A total of 1478 testees, including 342 pilots and 1136 non-pilots, are chosen randomly from an airline company and are randomly classified into a training group and a test group before performing Cattell’s 16 personality factor test. The 16 factors in the test are learnt by a support vector machine (SVM), and the learning results are analyzed. Results Five factors are used as eigenvectors for the classification. The classifier that is constructed based on linear SVM achieves a 78% average accuracy in the cross-validation. Conclusion The SVM-based classifier has high reliability and effectiveness.