Ensemble Usage for Classification of EEG Signals A Review with Comparison
| Autor: | Umair Muneer Butt, Sultan Zia, Sukumar Letchmunan, Zaib Unnisa, Sadaf Ilyas |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
business.industry
Computer science 05 social sciences Decision tree Pattern recognition Logistic regression Ensemble learning 050105 experimental psychology Random forest Support vector machine Set (abstract data type) 03 medical and health sciences Range (mathematics) ComputingMethodologies_PATTERNRECOGNITION 0302 clinical medicine 0501 psychology and cognitive sciences Artificial intelligence business 030217 neurology & neurosurgery |
| Zdroj: | Augmented Cognition. Theoretical and Technological Approaches ISBN: 9783030503529 HCI (16) |
| DOI: | 10.1007/978-3-030-50353-6_14 |
| Popis: | Significance: Ensemble learning is a robust and powerful approach to solve a variety of classification problems. Its usage has increased dramatically in recent years but not seen extensive application in EEG based Brain computer interface (BCI) problems. There is a wide range of classifiers which may not perform well, when used separately for classification problem but outperforms state-of-art algorithm when used as an ensemble, i.e., Long-short Term Memory (LSTM) is considered as best learning algorithms when time is embedded in input but not shown outstanding performance when used individually for EEG classification. On the contrary, it provided real good results when used as an ensemble. Aim: Aim of this study is how EEG signals can be classified using Ensembles methods; its importance and usage are described with experimental results. Approach: The approach that is being used is, combining different classifiers, i.e. Support vector machines, Decision Trees, Random Forest, Long Short Term Memory (LSTM), Logistic Regression (LR) and see which classifiers work best with which ensemble technique for EEG classification’s problem. Datasets: Datasets are taken from well-known data resources, Kaggle, EEG data set of confused students. The second dataset is taken from GitHub having EEG signals with timestamps according to events, i.e., sound, light, etc. According to our results, the LSTM- ensemble outperformed all other algorithms in the case where time is embedded in data. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|