Where academic tradition
meets the exciting future

You are here: TUCS > PUBLICATIONS > Publication Search > Evaluation of a Fuzzy C-Means ...

Evaluation of a Fuzzy C-Means Model in Currency Crisis Prediction

Dorina Marghescu, Shuhua Liu, Peter Sarlin, Evaluation of a Fuzzy C-Means Model in Currency Crisis Prediction. TUCS Technical Reports 970, Turku Centre for Computer Science, 2010.

Abstract:

In this paper, we analyze financial crisis data using the Fuzzy C-means (FCM) clustering technique. The aim is to investigate the extent to which this method is useful in predicting currency crises in emerging economies. The focus is on the currency crises that took place in Asia during 1997. We analyze a dataset that concern 23 developing countries all over the world and five variables that have been experimented in other similar studies. Thus, we build, using the FCM technique, an early warning system for forecasting the Asian crises based on historical data. The approach consists of clustering the data using the FCM technique and then classifying the clusters into early warning and tranquil clusters. We compare the prediction results of the model with those obtained by applying probit analysis. This study shows that the training accuracy of a FCM model is better than that of the probit model, while the prediction accuracy calculated for the test datasets is slightly worse. However, the FCM model has a good explanatory feature of different currency crises, that is, diverse crises over different periods of time and in different countries are assigned to various clusters whose centers can be regarded as representative conditions for financial instability. Therefore, we conclude that the high classification accuracy in training, coupled with the intuitive explanations of different crisis situations are indicators that a model based on the FCM technique could be useful for analyzing and predicting financial instability. However, future work is intended to improve the current model with respect to its predictive performance.

Files:

Full publication in PDF-format

BibTeX entry:

@TECHREPORT{tMaLiSa10a,
  title = {Evaluation of a Fuzzy C-Means Model in Currency Crisis Prediction},
  author = {Marghescu, Dorina and Liu, Shuhua and Sarlin, Peter},
  number = {970},
  series = {TUCS Technical Reports},
  publisher = {Turku Centre for Computer Science},
  year = {2010},
  keywords = {Fuzzy C-means, probit analysis, currency crisis, financial instability, early warning signals, prediction, evaluation},
  ISBN = {978-952-12-2423-2},
}

Belongs to TUCS Research Unit(s): Data Mining and Knowledge Management Laboratory

Edit publication