Clustering datasets with missing values using modified K-medoids algorithm /
Marbas, Ivan Art F.
Clustering datasets with missing values using modified K-medoids algorithm / Ivan Art F. Marbas. - 2008 - 61 leaves.
Thesis (BS Applied Mathematics) -- University of the Philippines Mindanao, 2008
A modification was done to the Euclidean distance to compute distance for incomplete data points, at the same time flagging them so that the algorithm will avoid choosing them as cluster medoids. This resulted to the Modified K-medoids clustering algorithm applied with pre-processing methods, namely, Case Deletion, Mean Imputation and K-nearest Neighbor Imputation, in clustering incomplete datasets, it showed that the proposed algorithm performs only second best to K-nearest Neighbor Imputation. The comparison was made using incomplete datasets generated from the Iris and Bupa dataset with different missing value occurrences and degradation levels. Though only second best, the production of cluster medoids with no missing values is unique to the modification. Thus, the Modified K-medoids clustering algorithm is more advantageous.
Clustering.
Incomplete datasets.
K-medoids.
Modified Euclidiean distance.
Datasets.
Euclidean distance.
Undergraduate Thesis --AMAT200,
Clustering datasets with missing values using modified K-medoids algorithm / Ivan Art F. Marbas. - 2008 - 61 leaves.
Thesis (BS Applied Mathematics) -- University of the Philippines Mindanao, 2008
A modification was done to the Euclidean distance to compute distance for incomplete data points, at the same time flagging them so that the algorithm will avoid choosing them as cluster medoids. This resulted to the Modified K-medoids clustering algorithm applied with pre-processing methods, namely, Case Deletion, Mean Imputation and K-nearest Neighbor Imputation, in clustering incomplete datasets, it showed that the proposed algorithm performs only second best to K-nearest Neighbor Imputation. The comparison was made using incomplete datasets generated from the Iris and Bupa dataset with different missing value occurrences and degradation levels. Though only second best, the production of cluster medoids with no missing values is unique to the modification. Thus, the Modified K-medoids clustering algorithm is more advantageous.
Clustering.
Incomplete datasets.
K-medoids.
Modified Euclidiean distance.
Datasets.
Euclidean distance.
Undergraduate Thesis --AMAT200,