Hybrid clustering technique of PCA-SM-GHSOM for abnormal and normal classification with quarterly financial ratios of listed TCM company sector
by Ruicheng Yang
International Journal of Simulation and Process Modelling (IJSPM), Vol. 11, No. 3/4, 2016

Abstract: This paper provides a hybrid technique PCA-SM-GHSOM for clustering the quarterly financial data into normal and abnormal groups. For evaluating the performance of this hybrid method, we give some empirical analysis for the listed traditional Chinese medicine (TCM) companies. Three stages are proposed for the clustering experiment. Firstly, we use the PCA method to reduce the high dimensions of financial ratios into low dimensions. Secondly, we adopt cosine similarity method to select three best matching companies in the same TCM sector, and further get the deviation dataset of the considered company. Finally, we put the deviation dataset into the GHSOM system and get the clustering results for training dataset and testing dataset respectively. Furthermore, we give some comparisons with other different techniques of single GHSOM, PCA-GHSOM and SM-GHSOM, and find that the proposed hybrid technique can improve significantly the accuracy for clustering the financial data into normal and abnormal groups.

Online publication date: Mon, 22-Aug-2016

The full text of this article is only available to individual subscribers or to users at subscribing institutions.

 
Existing subscribers:
Go to Inderscience Online Journals to access the Full Text of this article.

Pay per view:
If you are not a subscriber and you just want to read the full contents of this article, buy online access here.

Complimentary Subscribers, Editors or Members of the Editorial Board of the International Journal of Simulation and Process Modelling (IJSPM):
Login with your Inderscience username and password:

    Username:        Password:         

Forgotten your password?


Want to subscribe?
A subscription gives you complete access to all articles in the current issue, as well as to all articles in the previous three years (where applicable). See our Orders page to subscribe.

If you still need assistance, please email subs@inderscience.com