A Bootstrap Evaluation of the Effect of Data Splitting on Financial Time Series
8 Pages Posted: 22 Jan 1997
There are 2 versions of this paper
A Bootstrap Evaluation of the Effect of Data Splitting on Financial Time Series
A Bootstrap Evaluation of the Effect of Data Splitting on Financial Time Series
Date Written: December 1996
Abstract
This article exposes problems of the commonly used technique of splitting the available data into training, validation, and test sets that are held fixed, warns about drawing too strong conclusions from such static splits, and shows potential pitfalls of ignoring variability across splits. Using a bootstrap or resampling method, we compare the uncertainty in the solution stemming from the data splitting with neural network specific uncertainties (parameter initialization, choice of number of hidden units, etc.). We present two results on data from the New York Stock Exchange. First, the variation due to different resamplings is significantly larger than the variation due to different network conditions. This result implies that it is important to not over-interpret a model, or an ensemble of models, estimated on one specific split of the data. Second, on each split, the neural network solution with early stopping is very close to a linear model; no significant nonlinearities are extracted.
JEL Classification: G1, C5
Suggested Citation: Suggested Citation
Do you have negative results from your research you’d like to share?
Recommended Papers
-
Predicting Daily Probability Distributions of S&P500 Returns
By Andreas Weigend and Shanming Shi
-
A First Application of Independent Component Analysis to Extracting Structure from Stock Returns
By Andrew D. Back and Andreas Weigend
-
Modeling Volatility Using State Space Models
By Jens Timmer and Andreas Weigend