Exploiting Hierarchy in Text Categorization
Information Retrieval
Direct Domain Knowledge Inclusion in the PA3 Rule Induction Algorithm
PAKDD '01 Proceedings of the 5th Pacific-Asia Conference on Knowledge Discovery and Data Mining
Adequacy of training data for evolutionary mining of trading rules
Decision Support Systems - Special issue: Data mining for financial decision making
A hybrid genetic-neural architecture for stock indexes forecasting
Information Sciences: an International Journal - Special issue: Computational intelligence in economics and finance
Using the distribution of performance for studying statistical NLP systems and corpora
ELDS '01 Proceedings of the workshop on Evaluation for Language and Dialogue Systems - Volume 9
On the construction of a nonlinear recursive predictor
Journal of Computational and Applied Mathematics - Special issue: International conference on mathematics and its application
Constraint-based sequential pattern mining: the consideration of recency and compactness
Decision Support Systems
A neural network with a case based dynamic window for stock trading prediction
Expert Systems with Applications: An International Journal
On the construction of a nonlinear recursive predictor
Journal of Computational and Applied Mathematics
Hi-index | 0.01 |
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