Comparison of dimensional reduction using the Singular Value Decomposition Algorithm and the Self Organizing Map Algorithm in clustering result of text documents

Joint Conference on Green Engineering Technology & Applied Computing 2019

工业技术（总论）;计算机科学

Ihsan Jambak, Muhammad^1 ; Ikrom Izzuddin Jambak, Ahmad^1

Faculty of Computer Science, Sriwijaya University, Palembang, Indonesia^1

关键词: Alternative algorithms;    Artificial neural network models;    Dimensional reduction;    Feature extraction methods;    Feature selection methods;    Internal representation;    Self Organizing Map algorithm;    Singular value decomposition algorithms;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/551/1/012046/pdf
DOI  :  10.1088/1757-899X/551/1/012046

来源: IOP

【 摘 要 】

Dimension reduction has two methods namely feature selection and feature extraction. Dimension reduction using the feature selection method has a better influence than the feature extraction method on the cluster results. However, there is still a need for feature extraction methods to reduce dimensions. For this reason, an alternative algorithm is needed from the feature extraction method. Self Organizing Map (SOM) is one of the artificial neural network models that has a special nature that is effectively able to create spatial internal representations of input data, or in general to create smaller data dimensions. This research was examining the capability of SOM compared to Singular Value Decomposition (SVD) in reducing data dimension of text documents before there were clustered by k-Means. Results show that SVD still better than SOM in cluster quality index but SOM faster than SVD in computation times.