A weighted SOM for classifying data with instance-varying importance

This paper presents a weighted self-organizing map (WSOM) that combines the advantages of the standard SOM paradigm with learning that accounts for instance-varying importance. While the learning of the classical batch SOM weights data by a neighborhood function, it is here augmented with a user-specified instance-specific importance weight for cost-sensitive classification. By focusing on instance-specific importance to the learning of a SOM, we take a perspective that goes beyond the common approach of incorporating a cost matrix into the objective function of a classifier. This paper provides evidence of the performance of the WSOM on standard benchmark and real-world data. We compare the WSOM with a classical SOM and a conventional statistical approach in two financial classification tasks: credit scoring and financial crisis prediction. The significance of instance-varying importance weights, and the performance of the WSOM, is confirmed by being superior in terms of cost-sensitive classifications.

[1]  Elena Kalotychou,et al.  Optimal design of early warning systems for sovereign debt crises , 2007 .

[2]  Bianca Zadrozny,et al.  Learning and making decisions when costs and probabilities are both unknown , 2001, KDD '01.

[3]  Jari Kangas.  Sample weighting when training self-organizing maps for image compression , 1995, Proceedings of 1995 IEEE Workshop on Neural Networks for Signal Processing.

[4]  Peter Sarlin,et al.  Visual tracking of the millennium development goals with a fuzzified self-organizing neural network , 2012, Int. J. Mach. Learn. Cybern..

[5]  Guilherme De A. Barreto,et al.  Time Series Prediction with the Self-Organizing Map: A Review , 2007, Perspectives of Neural-Symbolic Integration.

[6]  Teuvo Kohonen.  THE HYPERMAP ARCHITECTURE , 1991 .

[7]  David J. Hand,et al.  Mining the past to determine the future: Problems and possibilities , 2009 .

[8]  Teuvo Kohonen,et al.  Self-Organizing Maps , 2010 .

[9]  Dong Zhou,et al.  Translation techniques in cross-language information retrieval , 2012, CSUR.

[10]  Charles Elkan,et al.  The Foundations of Cost-Sensitive Learning , 2001, IJCAI.

[11]  John G. Taylor,et al.  The temporal Kohönen map , 1993, Neural Networks.

[12]  Peter Sarlin,et al.  Mapping the State of Financial Stability , 2011, SSRN Electronic Journal.

[13]  Peter Sarlin,et al.  Self-organizing time map: An abstraction of temporal multivariate patterns , 2012, Neurocomputing.

[14]  Ana-María Fuertes,et al.  Early warning systems for sovereign debt crises: The role of heterogeneity , 2006, Comput. Stat. Data Anal..

[15]  Tom Fawcett,et al.  ROC graphs with instance-varying costs , 2006, Pattern Recognit. Lett..

[16]  Barbro Back,et al.  Combining visual customer segmentation and response modeling , 2014, Neural Computing and Applications.

[17]  Tom Fawcett.  PRIE: a system for generating rulelists to maximize ROC performance , 2008, Data Mining and Knowledge Discovery.

[18]  Kenneth Rogoff,et al.  Is the 2007 U.S. Sub-Prime Financial Crisis so Different? an International Historical Comparison , 2008 .

[19]  Uma Moorthy,et al.  Predicting Emerging Market Currency Crashes , 2002 .

[20]  Teuvo Kohonen,et al.  Things you haven't heard about the self-organizing map , 1993, IEEE International Conference on Neural Networks.

[21]  Tom Fawcett,et al.  Adaptive Fraud Detection , 1997, Data Mining and Knowledge Discovery.

[22]  Tuomas A. Peltonen,et al.  Assessing systemic risks and predicting systemic events , 2013 .

[23]  Christophe Hurlin,et al.  How to Evaluate an Early-Warning System: Toward a Unified Statistical Framework for Assessing Financial Crises Forecasting Methods , 2010 .

[24]  Esa Alhoniemi,et al.  Self-organizing map in Matlab: the SOM Toolbox , 1999 .

[25]  Peter Sarlin,et al.  On policymakers’ loss functions and the evaluation of early warning systems , 2013 .

[26]  Timo Honkela,et al.  WEBSOM - Self-organizing maps of document collections , 1998, Neurocomputing.

[27]  Sunil Vadera,et al.  A survey of cost-sensitive decision tree induction algorithms , 2013, CSUR.

[28]  Marie Cottrell,et al.  Advantages and drawbacks of the Batch Kohonen algorithm , 2002, ESANN.

[29]  Jong Beom Ra,et al.  Edge preserving vector quantization using self-organizing map based on adaptive learning , 1993, Proceedings of 1993 International Conference on Neural Networks (IJCNN-93-Nagoya, Japan).

[30]  Catherine Blake,et al.  UCI Repository of machine learning databases , 1998 .

[31]  Masanobu Taniguchi,et al.  Input dependent misclassification costs for cost-sensitive classifiers , 2000 .