Data Set A is a Pattern Matching Problem

Authors:
Jens Kohlmorgen;Klaus-Robert Müller
Affiliations:
GMD FIRST, Rudower Chaussee 5, 12489 Berlin, Germany;GMD FIRST, Rudower Chaussee 5, 12489 Berlin, Germany
Venue:
Neural Processing Letters
Year:
1998

Citing 1
Cited 1

Pruning with replacement on limited resource allocating networks by f-projections

Neural Computation

Applying LSTM to Time Series Predictable through Time-Window Approaches

ICANN '01 Proceedings of the International Conference on Artificial Neural Networks

Quantified Score

Hi-index	0.00

Visualization

Abstract

Several data sets have been proposed for benchmarking in time seriesprediction. A popular one is Data Set A from the Santa Fe Competition.This data set was the subject of analysis in many papers. In this note, itis shown that predicting the continuation of Data Set A is nothing elsethan a pattern matching problem. Looking at studies of this data set, itis remarkable that most of the very good forecasts of Data Set A usedupsampled training data. We explain why upsampling is crucial for thisdata set. Finally, it is demonstrated that simple pattern matchingperforms as good as sophisticated prediction methods on Data Set A.