Prepoznavanje i analiza spola prema akustičnim karakteristikama glasa i govora

master's thesis

• Author: Marina Bišćan
• Mentor: Anamarija Jazbec (mentor)
• Committee member: Anamarija Jazbec (predsjednik povjerenstva)
• Committee member: Bojan Basrak (član povjerenstva)
• Committee member: Miljenko Huzak (član povjerenstva)
• Committee member: Zvonimir Tutek (član povjerenstva)
• Granter: University of Zagreb; Faculty of Science; (Department of Mathematics); Zagreb
• Defense date and country: 2017-09-27, Croatia
• Scientific / art field, discipline and subdiscipline: NATURAL SCIENCES; Mathematics

Abstract

U ovom radu objašnjeni su glavni koncepti logističke regresije i klasifikacijskog stabla. Nakon kratkog uvoda i teorijske podloge, koristeći logističku regresiju i klasifikacijsko stablo, izvršili smo statističko predviđanje spola govornika prema akustičnim karakteristikama glasa i govora. Statistička analiza podataka izrađena je u statističkom programskom paketu SAS Studio. U primjeni je često varijabla od interesa diskretna, odnosno varijabla koja poprima vrijednosti iz prebrojivog skupa mogućih vrijednosti. Model logističke regresije se najčešće koristi za analizu upravo takvih podataka. Logistička regresija se koristi u različitim područjima, uključujući strojno učenje, najčešće u medicini i društvenim znanostima. Stabla odlučivanja spadaju u pristupe prediktivnih modela koji se koriste u statistici, u rudarenju podataka i strojnom učenju, te je njihova prednost jednostavnost i razumljivost metode. Metoda klasifikacijskih stabala temelji se na binarnom stablu, odnosno ukoliko je ciljna značajka diskretan skup vrijednosti. Obje metode pokazale su se uspješnima u predviđanju spola govornika.

Abstract (english)

This thesis explains the main concepts of logistic regression and classification trees and demonstrates their application. After a brief introduction and a theoretical background, using logistic regression and classification tree, an estimation of the speaker’s gender based on the acoustic characteristics of their voice and speech was performed. The statistical analysis of the data was made in the statistical program package SAS Studio ([1]). In applications the variable of interest is often discrete, that is a variable that receives values from a countable set of possible values. The logistic regression is most often used to analyze such data. Logistic regression is used in various fields, including machine learning, many medical fields, and social sciences. Decisions trees belong to the predictive modeling approaches used in statistics, data mining and machine learning. One of their main advantages is simplicity. The method of classification trees is based on a binary tree, i.e. if the target feature is a discrete set of values. Both methods have proven to be successful in estimating the speaker’s gender.

• Keywords: logistička regresija; klasifikacijsko stablo; statističko predviđanje; statistička analiza; SAS Studio; prediktivni modeli; rudarenje podataka; strojno učenje
• Keywords (english): logistic regression; classification tree; estimation; statistical analysis; SAS Studio; predictive modeling; data mining; machine learning
• Language: croatian
• URN:NBN: urn:nbn:hr:217:469992
• Study programme: Title: Mathematical Statistics; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra matematike (magistar/magistra matematike)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2019-03-27 12:44:45