The Model Edit Page in the Board Foresight application allows users to manage and refine their forecasting models. This page is organized into the following tabs:
Model Tab
This section enables users to make changes and tune the model's parameters. It includes options to lock/unlock the model, change the algorithm, rename it, and manage associated indicators. Users can also define the primary variable (the dataset being forecasted), set a target dataset (an alternative forecast), and adjust regression settings. Left-hand side of this tab is for model settings and bulk actions.
Forecasted: The primary indicator being forecasted by your Model.
Target: The target is used to calculate the risk section on the model output page and is used to compare an internal customer forecast with a forecast that is created on Foresight.
Tools
Add Indicator: Add an indicator to your Model for further analysis.
Condense List: Reduce the view of your graphs from visual charts to statistics.
Set Best Correlation Coefficient: Adjust the lead times of indicators to optimize based on the strongest correlation coefficient with the primary variable.
Set Best R-Squared: Adjust lead times according to the strongest R-squared value, indicating the best fit for regression analysis.
Set Best Procylic: Optimize indicator lead times based on pro-cyclical indicators, which move in the same direction as the economic cycle.
Set Best Countercylic: Set lead times for counter-cyclical indicators, which move in the opposite direction of the economic cycle.
Reset: Reset all secondary indicator offsets to 0.
Manage Indicators: In a pop-up modal, edit, lock, exclude, or delete indicators from the model. Include a screenshot for this modal.
Show Excluded: Display indicators that have been excluded from the model, allowing you to reconsider their use.
All Zero Offset Indicators Included in All Segments: Include all indicators that are set to zero offset in every segment of the model for comprehensive analysis.
Log Transform All Variables: Apply a logarithmic transformation to all model variables to stabilize variance and normalize data for regression analysis.
Model Settings
Regression Start Date: Set the beginning point for regression analysis, trimming any data before this date.
Analysis As Of Date: Controls the date range of the data used for generating forecasts. If set, it trims the end of the primary variable’s data and adjusts other variables accordingly.
Target Horizon: Dictates when to stop the forecast visualizations, controlling how much forecast data is displayed in charts and tables, i.e. 12 months. This can also be done based on a date, i.e. 01/01/2020.
Model State: Indicates the current status of the model:
Draft
Archive
Production
Expired
Factor Forecast
Calculation: Displays the type of calculation applied to the model, such as:
None
Period over Period
Year over Year
Toggles
Points: Controls the visibility of point symbols that represent individual data points on the forecast chart.
Forecast: Calculated data forecast.
Seasonality: Adjusts the data to account for seasonal patterns.
Adjustment Factors: Allows the application of factors to modify the primary variable for better handling of anomalies in the data caused by events like holidays or weather.
Right-hand Side Functionality
This includes the charts, offsets, indicator management.
Details: View in-depth information about the selected indicator.
Transform: Apply transformation techniques to the indicator.
Lock: Lock your indicator to prevent any changes to the indicator.
Exclude: Hide the indicator from view, so it will not be used when generating models.
Revert Forecast: Revert forecast to a previous state.
Indicator Display Information: View the metadata and statistics for each indicator.
Remove: Remove your indicator from the Model.
Do Not Include In All Segments: Prevent the indicator from being included in all segments of the Model.
Remove for Independent Variables: Remove the indicator from being considered as an independent variable.
Results Tab
This tab provides detailed performance measures and statistics of the model. It displays metrics related to forecasting accuracy, such as average single-period error and average aggregate error.
The results are generated based on the model's performance against historical data, allowing users to assess the model's effectiveness over time.
Analysis As Of Date: View the set date that your data will be considered from.
Tools
Manage Snapshots: View the captured states of your models.
Manage Locked Models: View and alter your locked models such as renaming and deleting your models.
Forecast: View forecast charts displaying predicted data such as raw forecasts, period over period forecasts, year over year forecasts, etc.
Accuracy: View accuracy calculations including single period error and aggregate period error, showcasing how well the forecast performs against the actual data.
Consistency: View the consistency of forecast results over different time periods.
Model + Standard Coefficient Toggle: Switch between viewing the model outputs in terms of standard coefficients or raw coefficients, providing different perspectives on the model results.
Raw + Log Toggle: Switch between displaying the data in its raw form or applying a logarithmic transformation to the indicators in the model.
Contribution: View how much each underlying indicator contributed to the total forecast for the period.
Component Contribution Report: A display of how much each underlying indicator contributed to the total forecast for a specific period.
Relative Importance: A measure based on the amount of historical variance in our dependent variable that each indicator explains over the shared time horizon.
Frequency Drop-Down Menu: This menu allows users to filter the data displayed in the Waterfall Chart based on different time frequencies (e.g., monthly, quarterly, semi-annual, and annual).
Waterfall Chart: The representation of how different elements affect the model's output, allowing users to see both positive and negative contributions clearly.
Start Date: The Waterfall Start Date defines the point at which the Waterfall Chart begins displaying contributions. Adjusting this date can help focus on specific time frames for better analysis of changes and contributions.
Stats: View statistical measures and performance metrics of the model including reliability and predictive capabilities.
Show Advanced Statistics Toggle: This toggle allows users to access additional statistical measures such as:
T-Value: A statistic used to determine the significance of individual explanatory variables in the regression model. It represents the ratio of the coefficient estimate to its standard error. A higher absolute T-value indicates a more significant effect of the explanatory variable on the primary variable.
Coefficient After Dropping Influential Outliers: The estimate of the relationship between an explanatory variable and the primary variable, calculated after removing data points that have a disproportionate impact on the model’s results (influential outliers). This coefficient provides a clearer understanding of the relationship without the distortion caused by these outliers.
95% Confident Interval in Coefficients: The range within which the user can be 95% confident that the true population parameter (the coefficient) lies. It provides insight into the precision of the coefficient estimate; a narrower interval suggests a more precise estimate.
Delete: This option allows the user to delete the entire model from the project along with all associated items. This action is permanent and removes the model from the system.
Hide: This tool allows the user to hide specific indicators from the view. While an indicator is hidden, it will not be used to generate the model, allowing for a cleaner presentation or focus on specific data points without permanently removing them from the model.
Diagnostics: View insights into model quality improvements and frequency of indicator usage during the forecast.
Raw + Log Toggle: This toggle allows users to choose whether to display the diagnostics data using the raw data values or the logged values. When activated, it shows results based on the logarithmic transformation. This is particularly useful when modeling datasets that exhibit exponential growth or have skewed distributions, as log transformations can help stabilize variance and improve model performance.
Log Cloud: This feature represents the visualization of logged data points, providing a view of how data behaves under logarithmic transformation. It helps users assess the impact of the log transformation on their data when analyzing the diagnostics and can offer insights into the model's handling of the underlying patterns in the data.
Info: View the details the model parameters, including independent variables used in the model.
Messages: View notifications or alerts related to model performance or adjustments needed.
Lineage: Track the origin and transformations of the model data, allowing for better understanding of data flow.
Notes: Document relevant observations or comments regarding the model. The user can also include input from AI generated insights.
Overview: View the summarization of key results and metrics of the forecast model.
Health: View insights of the overall health or performance of the forecast model.
Test Tab
This section is used to run tests and simulations on the model to evaluate its predictive capabilities and robustness.
Save and Run: This option triggers the model to execute the latest changes, saving any modifications made prior to running the analysis.
Export Last Executed Test to CSV: This functionality allows users to export the results of the last executed test into a CSV file for further analysis or reporting.
Execution Drop-Down Menu: This menu provides various execution options for running tests, allowing users to select different parameters or scenarios for the test.
Analysis As of Date: The "As of" date lets users dictate the data through which the model calculations are made. If set, the model will only consider data available up to that specific date, tailoring results to that timeframe.
Run Execution with Indicator Versions from: This option allows users to run the execution with particular versions of indicators as of a specified date. This can be useful for back testing or analyzing model performance with historical data.
Runs Current Data: Execute the model using the most up-to-date data available within the application. This function initiates the model calculations based on the latest information, ensuring that outputs reflect any new data that has been uploaded or modified since the last execution.
Compare Tab
In this tab, users can compare different iterations of the model by using executions and snapshots to analyze changes and improvements.
Execution Snapshots: Users can compare different snapshots of the model’s execution over time.
Forecast Drop-Down Menu: Select the type of calculation applied to the model.