Modify Data Series Dialog: Predictions
The Modify Data Series Dialog allows you to analyze and change the properties associated with a data series. The Predictions page allows you to analyze and retrain prediction fields for the data series.
Ä Note: If no prediction fields have been created for the data series, the dialog will indicate that no predictions have been created.
& For help with predictions, see Predicting and Modeling Financial Data.
Comparing Predictions
The Predictions page can be used to easily compare the error, correlation, and accuracy of the predictions for a data series. The predictions can be grouped by desired field, desired type (signals and non-signals), or not at all.
¨ Group By
This value indicates how to group the display of the predictions.
· Desired Field
This setting indicates to group the predictions by their desired outputs. This is useful for determining the best prediction of a specific field.
· Desired Type
This setting indicates to group the predictions by whether the desired output is a signal or a non-signal. This is useful for determining the best prediction of a signal when multiple desired output signals are used.
· No Grouping
This setting indicates not to group the predictions.
¨ Analyze
This value indicates which portion of the training data to display an analysis for. Typically, the testing set is used to compare the effectiveness of predictions.
· Pre-Optimization Data
This setting indicates to analyze all of the data before the beginning of the optimization set for each prediction.
· Training / Cross-Validation / Accuracy Testing Set
This setting indicates to analyze the data in a particular set for each prediction.
· Post-Optimization Data
This setting indicates to analyze all of the data after the end of the optimization set for each prediction.
· All Data
This setting indicates to analyze the data for all of the data in the data series.
Ä Note: If a Custom Range is specified, this selection will be replaced with the most recent custom range.
p Custom Range…
This button displays the Custom Date Range Dialog, which allows you to select a specific date range for all predictions.
Ä Note: Selecting a custom range (or all data) is different from selecting an optimization range since it causes the same date range to be analyzed for all predictions.
The information in each of these columns can be sorted. To sort a column of information, double-click on the heading of that column. The values in that column will then be sorted from best-to-worst, or alphabetical order for field names.
The following information is listed.
p Samples
This is the number of samples (bars) included in the selected optimization set or date range.
p Error
For predictions that use change preprocessing, this value is the error of the predicted changes. For all other predictions, this value is the raw error of the prediction.
p Correl.
For predictions that use change preprocessing and signals, this value is the correlation of the postprocessed value. For all other predictions, this value is the raw correlation.
p Accuracy
For predicted signals, this is the percent of signals that were in the correct direction. For all other predictions, this value is the percent of time the predicted change was in the correct direction.
Ä Note: Information on each of these values is available in the help for the Modify Field Dialog: Training Analysis page.
Viewing a Detailed Analysis
To analyze a prediction field, select the field, then press the Analyze… button. This will display the Modify Field Dialog: Training Analysis page. You can also analyze a prediction field by double-clicking the field name in the selection tree.
While viewing the Modify Field Dialog, you can change any of the properties of the prediction by selecting a different property page. However, the analysis will not be updated to reflect those changes until the OK button is pressed and the field has been retrained and reanalyzed.
& For help using the selection tree, see the help for Selection Trees.
Retraining a Prediction
To retrain a prediction field, select the field, then press the Retrain button. Your action will be confirmed and the prediction will be retrained.
& For help using the selection tree, see the help for Selection Trees.
What Do I Do Next?
When you are finished analyzing the predictions and making any changes, press the Close button or select another property page.
Additional Property Pages
The Modify Data Series Dialog has the following property pages.
This page allows you to change the ticker and display names for the data series, as well as providing you with several quick actions to perform on the data series.
This page allows you to add new fields, as well as modify and delete existing fields in the data series.
This page allows you to adjust the trading settings associated with this data series.
This page allows you to analyze and retrain prediction fields for the data series.
This page allows you to compare and analyze signal fields for the data series.
This page allows you to manage previously saved chart and spreadsheet views for the data series.
& For help using property pages, see the help for Property Pages.
How Did I Get Here?
The Predictions property page appears when you select Analyze Predictions… from the context menu for a data series in the Portfolio View.