Resampling in neural networks with application to financial time series
Bibliographic record
Abstract
Neural networks provide powerful data analysis tools to handle various types of nonlinearities in many areas, especially in the area of financial time series prediction. The computationally oriented 'jackknife' and 'bootstrap' neural network learning algorithms are developed in this thesis to forecast noisy financial time series, in particular, the spot Canada/US foreign exchange rate. The daily traded foreign exchange rate (FX) is influenced by many factors. In the case of Canada and the US, capital will flow into the country with the preferable yield, directly influencing the spot Canada/US FX. Therefore, one set of variables always on the traders monitor, is the interest rate spreads between comparable secure and highly liquid assets; in particular, the short term interest rate spread. In this thesis, nonlinear transfer function models between the short term interest rate spread and the spot Canada/US FX are studied by using multi-layer feed-forward neural networks, in conjunction with 'back-propagation' ( BP) learning and associated statistical re-sampling methods of the ' jackknife' and the 'bootstrap'. Prediction of the spot Canada/US FX will be the focus. The basic modeling strategy is to build a forecasting model which satisfies both the "trader experience criteria" and the underlying mathematical/statistical structure. Several neural network predictive models are proposed using multi-layer feed-forward neural network architectures. In addition, the stability property of the nonlinear transfer predictive model is studied by using local stability analysis. Second, a comparative pre-test of the neural network model is constructed to evaluate the network performance and to select the 'best' model for further study. All of the testing models give us about 55%-60% accuracy of the directional forecast on the "out-of-sample test set". Comparing with the linear predictive model, a 2% to 5% gain is obtained by using nonlinear neural network models. Consequently, the separate neural network model explores the nonlinear structure between the spot Canada/US FX and short term interest rate spread, especially during the time period of negative interest rate spread. It also captures a corrective mean reversion when the Canadian dollar is under-valued or over-valued in the market. Furthermore, the comparative pre-test demonstrates the impact changes in the interest rate spread has on changes in the spot FX. Statistical data-based re-sampling methods such as 'jackknife ' and 'bootstrap' are studied in the case of cross-validation BP learning. Two 'grouped jackknife' and two ' bootstrap' cross-validation learning algorithms are proposed, using parametric and non-parametric nonlinear modeling methodologies. The results show that both 'grouped jackknife' and 'bootstrap' learning algorithms lead to robust and reliable forecasts along with the large amount of statistical information of the previous knowledge. (Abstract shortened by UMI.)
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.006 | 0.020 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.002 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.002 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".