Banca de QUALIFICAÇÃO: Fernando Narciso Bertolaccini de Souza

Uma banca de QUALIFICAÇÃO de MESTRADO foi cadastrada pelo programa.
STUDENT : Fernando Narciso Bertolaccini de Souza
DATE: 26/02/2024
TIME: 15:00
LOCAL: Plataforma Teams
TITLE:

Predictive Model for Military Evasion Air Force Command Developed by Application of Learning Algorithms Machine

KEY WORDS:

Brazilian Air Force, Military Evasion, Machine Learning, Predictive Model, CRISP-DM, Evasion Prediction, Classification Techniques, Cross-Validation.

PAGES: 27
BIG AREA: Ciências Exatas e da Terra
AREA: Ciência da Computação
SUBÁREA: Metodologia e Técnicas da Computação
SPECIALTY: Sistemas de Informação
SUMMARY:

This dissertation addresses the creation of a predictive model to identify military personnel attrition in the Brazilian Air Force (Comando da Aeronáutica (COMAER)), as well as the prediction of attrition risk rate using machine learning techniques. The personnel management in COMAER faces the challenge of forecasting the number of active military personnel in each cadre and rank over the years, considering that some military personnel become inactive before the expected time for their careers, impacting higher-level positions that may remain unoccupied in the future. The study proposes the use of the CRossIndustry Standard Process for Data Mining (CRISP-DM) model to develop and implement the classification and prediction model for attrition. The model will be developed based on historical data collected from the Sistema de Gerenciamento de Pessoal (SIGPES), containing information about military personnel who have left and those who have remained in the institution. The data will undergo a preprocessing process to remove redundant and missing information and balance the class distribution between training and validation data. Supervised learning techniques, such as Logistic Regression, Decision Trees, SupportVector Machine (SVM), and K-Nearest Neighbors (k-NN), will be applied to train the model. The evaluation will be performed using metrics such as Accuracy, Precision, Specificity, Sensitivity, and F1-Score, ensuring the robustness of the model and avoiding overfitting. The implementation of this predictive model will enable the adoption of strategic actions, such as the creation of personnel retention policies, to minimize military attrition and ensure the efficiency and readiness of the Brazilian Armed Forces.

COMMITTEE MEMBERS:
Presidente - 2866456 - GLADSTON LUIZ DA SILVA
Interno - 3000020 - GUILHERME SOUZA RODRIGUES
Interno - 402520 - MARCELO LADEIRA
Externa à Instituição - ANA PAULA BERNARDI DA SILVA - UCB