Training Analysis Page: Overview

The Modify Field Dialog allows you to analyze and modify the properties of a field. The Training Analysis page allows you to view an analysis of the prediction results versus the desired values.

&  For help with predictions, see Predicting and Modeling Financial Data.

Overview Sub-Page Data

This sub-page displays an overview of the training. The following information is provided.

p    Testing Summary

These values are a summary of the testing values that are displayed on the other sub-pages. For predictions of fields that are not signals, these values are displayed for the actual value and the percent change value. For predictions of signal fields, these values are displayed for the raw and postprocessed values.

p    Normalized Error

These values are the normalized error of the difference between the desired values and the predicted values normalized using the last known value. For more information on these values, see the help for the Normalized Error sub-page.

p    Correlation

These values are the statistical correlation between the desired values and the predicted values. For more information on these values, see the help for the Correlation sub-page.

p    Directional Accuracy

This value displays an analysis of the directional accuracy of the predicted values. For more information on this value, see the help for the Directional Accuracy sub-page.

Ä    Note: This value is not available for predicted (modeled) signals.

p    Accuracy (Non-Hold)

These values display an analysis of the directional accuracy of the predicted signals. For more information on these values, see the help for the Prediction Matrix sub-page.

Ä    Note: This value is only displayed for predicted signals.

p    Ranges

These values summarized the ranges of data used for the training, cross validation, and accuracy testing set.

p    Starting Date / Ending Date

These values display the starting and ending date for this subset.

Ä    Note: These date ranges can be adjusted on the Training Optimization Settings Dialog.

p    Overall Samples

These values indicate the total number of samples that were available in this subset.

Ä    Note: These sample sizes will often be different from the approximate sizes displayed on the Training Optimization Settings Dialog. This is because the approximate sizes do not take into account lagged, null, or conditional data.

p    Changes Upward / Changes Downward

These values indicate the percentage of the overall samples in which the data changed upward or downward from the previous sample. This is useful for determining if a disproportionate number of samples changed in a given direction, which may cause the prediction to specialize on the predominant direction.

p    Positive Signals / Negative Signals

These values indicate the percentage of the overall samples in which the signal was above or below the exit threshold. This is useful for determining if a disproportionate number of samples were for a certain signal type, which may cause the prediction to specialize on the predominant signal.

Ä    Note: These percentages do not include hold signals.

Overview Sub-Page Analysis

The analysis presented on this page is based on the values displayed in the testing summary. TradingSolutions will try to find the best characteristic of the summary and highlight it. This can be useful for determining how a model can best be used.

The evaluation of models is very abstract. Models that are classified as "excellent" or "good" may be good at mirroring the desired data, but may not predict the values in a way that is useful in the way that was intended. Similarly, models that are classified as "poor" may produce values that are still useful. The analysis text is simply provided as a starting point for evaluating the model as a whole.

If a problem occurred during the training or calculation phase, the analysis will be replaced with a description of the error. A summary of these error messages is displayed on the help for the Modify Field Dialog: Training Analysis page.

Overview Sub-Page Actions

To retrain the network immediately, press the Retrain Prediction button. To resume the existing training from where if left off, press the Resume Training button.

Ä    Note: If any settings relating to the training of the prediction have been changed since the last training, the resume button will not be available.

To view the postprocessing values associated with a predicted signal, press the View Postprocessing… button. This will display the Signal Postprocessing Settings Dialog.

How Did I Get Here?

This is a sub-page of the Modify Field Dialog: Training Analysis page.