Deep Neural Network to Predict NDA Graduates Refusing to Become SDF Officers

Kazuhiro Seiwa, Keisuke Iwai, Takashi Mastubara, Takakazu Kurokawa


This paper proposes feed forward neural network architecture to predict NDA graduates refusing to become Self Defense Force officers on National Defense Academy of Japan. Using the actual student data including 24 items such as course grades and club activities, our deep neural network recognizes the pattern of the students. It will resolve the problem on personnel affairs. Through neural network architecture has some hyper parameters which need adjustment for each task, its simulation results showed the optimum parameters by several verifications. The result of this research shows it becomes possible to predict NDA graduates refusing to become SDF officers with an accuracy of 93.0%, and its F-measure equals to 0.31.


neural networks; random forests; support vector machine; machine learning

Full Text:



  • There are currently no refbacks.