Institutional Repository of Vilnius University
Konvoliucinių dirbtinių neuroninių tinklų taikymas statistinėje duomenų analizėje /
Abstract
dc:descriptionNowadays, amount of data generated by IT users is growing at a significant pace. To process all of the generated data, new ways statistical analysis automatization are created. One of these ways is machine learning. Probably, the most popular technnique used in machine learning are artificial neural networks. This paper looks into application of convolutional neural networks, one of the most widespread types of neural networks, in statistical analysis. The dataset for the expirement is the set of Vilnius public transportation lag from 01-01-2018 to 07-05-2018. After training the convolutional neural network with chosen dataset, influence of neural network’s parameters to neural network’s accuracy was analyzed. The experiment has yielded following results: accuracy is substantialy better than taking a random guess (29% vs 8%) or applying popular machine learning algorithms like Naive Bayes and Random Forset (about 7.8%), the maximum accuracy can be achieved with low numbers of layers (up to 3), hinge loss function is suitable for the chosen dataset, L2 regularization positively affects accuracy of the neural network, relu non-linearity function positively influences accuracy of the neural network.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2018
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Balkunas, Edvardas,
- Contributors dc:contributor
-
- Valaitis, Vytautas
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD29548704&prefLang=en_US