Use of value-based and motivational parameters with artificial intelligence technology to predict cadet maladjustment

Abstract

Listen

Translate article

The paper demonstrates the potential for using value-based and motivational parameters with artificial intelligence technology to predict cadet maladjustment. A retrospective cohort study was conducted. For 2013–2021, 734 cadets of the Navy Military Training and Research Center “Soviet Union Fleet Admiral N.G. Kuznetsov Naval Academy” were examined, 48 of them were diagnosed with maladjustment. Neural networks were used for mathematical modeling of maladjustment prediction. The study included 8 cycles of neural network training and 7 cycles of neural network model testing. As the actual material increases, the sensitivity of the model for predicting cadet maladjustment using neural networks increases: 30.MLP 16-7-2; 28.MLP 16-13-2; 30.MLP 16-22-2; 29.MLP 16-31-2; 42.MLP 16-39-2; 19.MLP 16-45-2; 16.MLP 16-48-2; 30.MLP 16-30-2 from 0.43 to 1.00 conventional units (y = 0.017x2 – 0.0647x + 0.4898, R² = 0.8264); specificity: from 0.96 to 1.00 conventional units (y = –0.002x2 + 0.0211x + 0.9462, R² = 0.8923); predictive value increased from 91.8% to 99.45% (y = –0.1477x2 + 2.3309x + + 90.238, R² = 0.9368). When the models were tested on new samples, the mean sensitivity was 0.45 conventional units with an increasing trend (y = 0.0207x2 – 0.1214x + 0.5271, R² = 0,6945), specificity: 0.97 conventional units (y = –0.0048x2 + + 0.0388x + 0.9086, R² = 0.772), predictive value: 92.6% (y = –0.4962x2 + 3.5402x + 88.447, R² = 0.6598). Therefore, the model for predicting cadet maladjustment using neural networks can identify cadets who will experience maladjustment with an accuracy of 32% to 72%, whereas no more than 6% of cadets without maladjustment will receive a false prediction. The predictive value of the model is close to the absolute accuracy of vocational aptitude prediction with reference values of 65%–70%. The predictive ability of the models tested in the study, ranging from 89.7% to 96.4%, confirms the high effectiveness of using neural networks to predict maladjustment. The value-based and motivational parameters of the cadets, combined with the use of neural networks to predict their maladjustment, create a highly effective artificial intelligence system. Such an approach can be used in medical and psychological support activities for military personnel at a military university for their optimal selection and support.