Note
Access to this page requires authorization. You can try signing in or changing directories.
Access to this page requires authorization. You can try changing directories.
While training the prediction model, you might come across the messages in this article that AI Builder might report. Messages are either errors or warnings. Each is represented by an icon.
| Message | Icon |
|---|---|
| Error |  |
| Warning |  |
When an error occurs, you can't continue until you resolve it. If the system is unable to correct a problem, it will show you an error.
Warnings are messages reported as informational. They don't stop you from proceeding. They warn you of possible performance issues when training the model.
Error: No AI Builder license
You need an AI Builder license to use this feature. Start or extend a trial or contact your admin to upgrade.
Cause
You either don't have an active AI Builder license, or the existing trial has expired.
Resolution
To use AI Builder models, make sure you have the AI Builder license assigned.
Error: Insufficient number of rows to train
The model needs at least 50 rows to train. <TableName> has only <ActualValue> rows. Add data or select another table.
Cause
The table that you've selected as the historical outcome doesn't have enough rows in it for the model to train itself to predict future outcomes.
Resolution
Add a minimum of 50 rows to the table. Use a minimum of 1,000 rows for better prediction results. More information: Prediction model prerequisites
Error: Insufficient historical outcome rows to train
The model needs at least 10 historical outcome rows of each outcome value to train. Add data or select another table.
Cause
The column that you selected to let AI Builder study the historical outcome doesn't have enough rows for each possible outcome. For example, in a Boolean field where the possible outcomes can be either true or false, there should be a minimum 10 rows of historical data where the outcome is set to True and another 10 rows set to False.
Resolution
Make sure you have 10 rows of each possible outcome value that you would like the model to predict. More information: Prediction model prerequisites
Warning: Add data to improve model performance
The model might produce better performance with optimum rows of 1,000 or more to train the model. Online Shopper Intention has <Actualcount> rows. Add data for better model performance.
Cause
The AI model found that the number of rows might not result in optimal model performance.
Resolution
We recommend 1,000 rows or more of historical data with outcomes to predict outcomes with a high level of accuracy. However, 50 rows is the minimum limit to process the prediction model.
Warning: Column might be dropped from training model
<TableName>.<ColumnName> might get dropped from training as it has a single value and does not contribute to training the model.
Cause
The AI model processes data in the columns related to the outcome that will influence the prediction. Of the various columns selected, it found that the specified column had only a single value across all rows in the table. Because of this, it won't impact the prediction and won't help training the model.
Resolution
Make sure that all columns selected as being related to the outcome column have multiple values in the column. This will help with the training of the model.
Warning: High ratio of missing values
<TableName>.<ColumnName> has a high ratio of missing values, greater than <ThresholdValue> percentage and might not contribute to train the model.
Cause
The AI model processes the data in the columns related to the outcome that will influence the prediction. Of the various columns selected, the model found that the specified column had data in few rows in the table. Since the data won't impact the prediction, it won't help training the model.
Resolution
Make sure the columns that are selected as being related to the outcome have data for them across most of the rows in historical data.
Warning: High percent correlation to the outcome column
<TableName>.<ColumnName> has <ThresholdValue> percent correlation <CorrelationName> with <OutcomeAttributeName> and model might suspect to cause target leak.
Cause
The AI model processes the data in the columns related to the outcome that will influence the prediction. Of the various columns selected, it found that the specified column has a high correlation with the outcome column, which might impact the prediction result. Because of this, it won't be included in training the model.
Resolution
Make sure the columns selected as being related to the outcome don't have a high correlation with the outcome column for a fair prediction.