Model glasača na predsjedničkim izborima baziran na metodama strojnog učenja

master's thesis

• Title (english): Voter Model in Crotian Presidential Elections Based on Machine Learning Methods
• Author: Mariela Markovinović
• Mentor: Dario Bojanjac (mentor)
• Committee member: Dario Bojanjac (predsjednik povjerenstva)
• Committee member: Sonja Grgić (član povjerenstva)
• Committee member: Tomislav Kos (član povjerenstva)
• Granter: University of Zagreb; Faculty of Electrical Engineering and Computing; Zagreb
• Defense date and country: 2020-07-14, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Computing

Abstract

U ovome radu prikazani su različiti algoritmi storojnog učenja. Analizirana je ovisnost točnosti modela o hiperparametrima kao i o odabranom skupu podataka za predviđanje. Skup podataka za analiziranje čine dvije ankete koje ispituju stavove glasača neposredno prije prvog i drugog kruga predsjedničkih izbora. Opisane su metode prikupljanja, pripreme i analize anketa. Prvi dio rezultata čine testovi šest algoritama, prikazana je ovisnost točnosti o različitim hiperparametrima. Drugi dio čine predviđanja distribucije glasova pomoću algoritma slučajne šume. Najbolji rezultati dobivaju se korištenjem algoritma slučajne šume, gdje je točnost predviđene distribucije glasova 85%.

Abstract (english)

This master's thesis presents tests of several different machine learning algorithms. The dependence of the accuracy of the model on the hyperparameters, as well as on the choice of the data-set used for the predictions, were analyzed. The data-set used in the analyses consists of two polls designed to examine voter attitudes just before the first and second rounds of the presidential election. Methods of collecting, preparing, and analyzing surveys are described in the master's thesis. The first part of the thesis presents the results of the tests of six different algorithms, expressed via the model accuracy dependence on different hyperparameters. The second part presents the distribution of votes as predicted using a random forest algorithm. The best results and highest accuracy are obtained using a random forest algorithm, where the accuracy of the predicted distribution of votes is 85%.

• Keywords: strojno učenje; neuronske mreže; logistička regresija; stablo odlučivanja; algoritam k-najbližih susjeda; slučajne šume; stroj potpornih vektora; predsjednički izbori
• Keywords (english): machine learning; neural networks; logistic regression; decision trees; K-nearest neighbors algorithm; random forest; support vector machine; presidential elections
• Language: croatian
• URN:NBN: urn:nbn:hr:168:136863
• Study programme: Title: Information and Communication Technology; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka informacijske i komunikacijske tehnologije (magistar/magistra inženjer/inženjerka informacijske i komunikacijske tehnologije)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Closed access
• Created on: 2023-01-13 13:00:43