Mollashahi, Mohsen (2021) Investigating the performance of k-Nearest Neighbor classifier based on two new distance functions in diagnosis of different cancers. Masters thesis, University of Zabol.
|
Text
Investigating the performance of k-Nearest Neighbor classifier based on two new distance functions in diagnosis of different cancers.pdf Download (158kB) | Preview |
Abstract
One of thesimplest and yet comparable efficiency and even in some optimal cases, is the complex and modern algorithms available in machine learning domain and especially the classification problem, the k - nearest neighbor (KNN) algorithm. The coreof this algorithm is directly dependent on the measurement of distance or similarity between the experimental sample and the training samples. Due to the frequency and variety of distance functions and topology and spatial geometry in which the data are located, especially in high dimensions, determining the optimal metric and testing new metrics is necessary to achieve maximum efficiency. In this study, the aim is to use the two new distance functions of Sobolev and Fisher for the first time in the KNN algorithm and to evaluate and compare their performance with existing distance functions for data, especially data related to the field of cancer diagnosis.
| Item Type: | Thesis (Masters) |
|---|---|
| Uncontrolled Keywords: | K-Nearest Neighbor, Classification, Distance Fisher, Sobolev, Cancer |
| Subjects: | Q Science > QH Natural history > QH301 Biology |
| Depositing User: | Mrs najmeh khajeh |
| Date Deposited: | 20 Oct 2021 05:26 |
| Last Modified: | 20 Oct 2021 05:26 |
| URI: | http://eprints.uoz.ac.ir/id/eprint/2825 |
Actions (login required)
 |
View Item |
