Kogemuslik masinõppealgoritmide võrdlus EEG andmete põhjal
Abstract
Selle töö eesmärgiks on võrrelda erinevaid masinõppealgoritme ning üritada leida nende hulgast parim EEG andmete klassifitseerimise jaoks. Selle saavutamiseks klassifitseeriti 10 inimese andmeid 10 masinõppealgoritmi poolt. Algoritme võrreldi kolmel viisil: esiteks võrreldi neid kolme erineva jõudlust iseloomustava näitaja alusel, teiseks kasutati klasteranalüüsi meetodeid ja dendrogramme ning viimaks kasutati selleks korrelatsioonimaatrikseid. Saadud võrdluse tulemused näitavad, et optimeerimata parameetrite korral on logistilise regressiooni mudel kõige efektiivsem algoritm EEG andmete klassifitseerimisel. Optimeeritud parameetrite korral on kõige efektiivsemaks algoritmiks juhumets.
The aim of this work is to compare different machine learning algorithms in an attempt to find the best one for classifying EEG data. In order to achieve this, the data from ten subjects were classified by ten machine learning algorithms. The algorithms were compared in three ways: Firstly, they were compared by using three performance metrics, secondly, by using clustergrams and lastly, by using corralation matrices. The results from the comparison show that the without parameter optimization, logistic regression model is the most efficient algorithm for classifying EEG data. However, with parameter optimization, random forest is the most efficient algorithm for classifying EEG data.
The aim of this work is to compare different machine learning algorithms in an attempt to find the best one for classifying EEG data. In order to achieve this, the data from ten subjects were classified by ten machine learning algorithms. The algorithms were compared in three ways: Firstly, they were compared by using three performance metrics, secondly, by using clustergrams and lastly, by using corralation matrices. The results from the comparison show that the without parameter optimization, logistic regression model is the most efficient algorithm for classifying EEG data. However, with parameter optimization, random forest is the most efficient algorithm for classifying EEG data.