Clustering of datasets with missing values using principal feature analysis as a feature selection tool / Frank B. Pelpinosas.
Material type:
TextLanguage: English Publication details: 2008Description: 51 leavesSubject(s): Dissertation note: Thesis (BS Applied Mathematics) -- University of the Philippines Mindanao, 2008 Summary: One of the most prevalent problemsAbstract: One of the most prevalent problems in clustering is the presence of redundant and irrelevant features, which could damage and misguide the clustering results of the data. Principal Feature Analysis is used as a filter feature selection tool to reduce highly dimensional datasets into smaller dimensions yet preserving the original structure of the data. The problem is worsened with the presence of missing values in the data. The study provides a comparison of the clustering results of the complete (base) datasets and imputed datasets using K-NN and mean imputation across three levels of degradation. The features retained by PFA were used to cluster the samples and were assessed using the Adjusted Rand Index. Results showed that PFA indeed had reduced the dimensions of the data. Principal Feature Analysis also can hardly drop some feature seven when charges in the levels of degradation appear. Both feature retention and cluster recovery were negatively affected by the number of missing values in the data in all the comparison.
| Cover image | Item type | Current library | Collection | Call number | Status | Date due | Barcode |
|---|---|---|---|---|---|---|---|
|
|
Thesis
|
University Library Theses | Room-Use Only | LG993.5 2008 A64 P44 (Browse shelf(Opens below)) | Not For Loan | 3UPML00012278 | |
|
|
Thesis
|
University Library Archives and Records | Preservation Copy | LG993.5 2008 A64 P44 (Browse shelf(Opens below)) | Not For Loan | 3UPML00032898 |
Thesis (BS Applied Mathematics) -- University of the Philippines Mindanao, 2008
One of the most prevalent problems
One of the most prevalent problems in clustering is the presence of redundant and irrelevant features, which could damage and misguide the clustering results of the data. Principal Feature Analysis is used as a filter feature selection tool to reduce highly dimensional datasets into smaller dimensions yet preserving the original structure of the data. The problem is worsened with the presence of missing values in the data. The study provides a comparison of the clustering results of the complete (base) datasets and imputed datasets using K-NN and mean imputation across three levels of degradation. The features retained by PFA were used to cluster the samples and were assessed using the Adjusted Rand Index. Results showed that PFA indeed had reduced the dimensions of the data. Principal Feature Analysis also can hardly drop some feature seven when charges in the levels of degradation appear. Both feature retention and cluster recovery were negatively affected by the number of missing values in the data in all the comparison.
Click on an image to view it in the image viewer
There are no comments on this title.