Raspoznavanje visokorizičnih mentalnih stanja pilota na temelju fizioloških značajki i strojnog učenja

master's thesis

• Title (english): Recognition of pilots' high-risk mental states based on physiological features and machine learning
• Author: Laura Palac
• Mentor: Siniša Popović (mentor)
• Committee member: Siniša Popović (predsjednik povjerenstva)
• Committee member: Alan Jović (član povjerenstva)
• Committee member: Nikolina Frid (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2025-02-27, Croatia

Abstract

Ljudska greška je jedan od glavnih uzroka zrakoplovnih nesreća, stoga pravovremeno prepoznavanje visokorizičnih mentalnih stanja pilota može značajno doprinijeti sigurnosti letova. Ovaj rad istražuje metode strojnog učenja za klasifikaciju kognitivno-emocionalnih stanja pilota reprezentiranih u javno dostupnom podatkovnom skupu, koristeći fiziološke značajke vezane za elektroencefalogram (EEG), elektrokardiogram (EKG), galvansku reakciju kože (GSR) i disanje. Implementirane su i analizirane tri... More metode strojnog učenja: stablo odluke, slučajna šuma i KNN. Koriste se dvije vrste klasifikacije: višeklasna i binarna. Učinkovitosti modela uvelike su ovisile o načinu podjele podatkovnog skupa na skup za trening i skup za test. Slučajna šuma bila je najbolji klasifikator u uvjetima višeklasne klasifikacije pri Leave-One-Subject-Out (LOSO) podjeli, a KNN u uvjetima binarne klasifikacije pri LOSO podjeli te binarne i višeklasne klasifikacije pri stratificiranoj nasumičnoj podjeli skupa podataka. Less

Abstract (english)

Human error is one of the leading causes of aviation accidents; therefore, timely recognition of high-risk mental states in pilots can significantly contribute to flight safety. This study explores machine learning methods for classifying pilots’ cognitive-emotional states represented in a publicly available dataset, using physiological features related to electroencephalography (EEG), electrocardiography (ECG), galvanic skin response (GSR), and respiration. Three machine learning methods were... More implemented and analyzed: decision tree, random forest, and KNN. Two types of classification were used: multiclass and binary. The effectiveness of the models largely depended on the method used for splitting the dataset into training and test sets. Random forest was the best classifier in case of multi-class classification under the Leave-One-Subject-Out (LOSO) split, while KNN performed best in binary classification under the LOSO split, as well as in both binary and multi-class classification under the stratified random split of the dataset. Less

• Keywords: pilot; mentalno stanje; fiziološke značajke; senzor; strojno učenje; simulator leta; pažnja
• Keywords (english): pilot; mental state; physiological features; sensor; machine learning; flight simulator; attention
• Language: croatian
• URN:NBN: urn:nbn:hr:168:408173
• Study programme: Title: Computing; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka računarstva (magistar/magistra inženjer/inženjerka računarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2025-03-12 11:15:07