Javascript must be enabled for the correct page display

Comparing Advanced Machine Learning Techniques for Classification Problems

Dijk, R. van (2016) Comparing Advanced Machine Learning Techniques for Classification Problems. Bachelor's Thesis, Artificial Intelligence.

AI_BA_2016_RUBENVANDIJK.pdf - Published Version

Download (672kB) | Preview
[img] Text
Toestemming.pdf - Other
Restricted to Backend only

Download (532kB)


This research compares (and combines) multiple advanced machine learning methods: the multi-layer support vector machine (ML-SVM), extreme gradient boosting (XGB), principal component analysis (PCA) and ensemble learning methods (bagging and stacking). These methods are compared on a large dataset for classification. On this dataset, XGB combined with an ensemble method achieved the highest accuracy. The eigenvectors obtained using PCA did not contribute to any promising results for XGB. Combining the dataset with principal components also did not contribute to promising results for XGB. Training the ML-SVM is a time consuming computation, especially on large datasets. The ML-SVM uses particle swarm optimization (PSO) in order to optimize its algorithm. In this research the complexity of the PSO on the ML-SVM is reduced by decreasing the search space of the PSO. These results are compared to the original results of the ML-SVM and the overall accuracy of the less parameter tuned ML-SVM was slightly less than the original, but higher than a single SVM. Also XGB is compared to these results: XGB performed worse than the ML-SVM on smaller datasets.

Item Type: Thesis (Bachelor's Thesis)
Degree programme: Artificial Intelligence
Thesis type: Bachelor's Thesis
Language: English
Date Deposited: 15 Feb 2018 08:14
Last Modified: 15 Feb 2018 08:14

Actions (login required)

View Item View Item