how to make prediction training table
Hello,
I would like to train a prediction model with a data table, but it returns error like this:
so my question: what is the standard table format to train a prediction model for single choice field?
Best Answers
-
@SHI_172837 The Predictive Models help article should be useful to you.
For a table to be used in a predictive model, it must have at least 1 field to use as the target for prediction that is one of these field types:
- Single Choice
- True or False
- Integer
- Decimal
The table should also have additional fields of any type that can be used as factors in making the prediction.
Based on your question, it sounds like you believe that table has a Single Choice field, but I'm guessing that perhaps it's currently designated as a text field.
To view your data table:
1. Open the Apps icon on the top-right side, then select Data.
2. Search for your table and open it.You may need to select the 3-dots icon in a column and select Edit field to update the field type.
Sometimes, it's best to start with a new table. To import historical data from an Excel spreadsheet:
1. Open the Apps icon on the top-right side, then select Data.
2. Select Create a Data Table and follow the prompts to import it.5 -
@Jeff_146001, I would like to further improve the accuracy of the prediction model, but I am not sure how exact to carry out the 2nd step in the "Guided Implementation":
I tried to manual select the prediction target field several times with below manual task assignment, but the prediction output did not seam to improve.
what is the missing part?0 -
Hi, @SHI_172837.
When using a master table where a row is added for each run of a workflow, the predictive model is automatically trained as your workflow runs.
When using an imported table (i.e. a .csv or .xlsx file that you uploaded in the Data section), to continually train the model, you'll need a Tables: Add a row step at the end of your sequence to add the new data.
As you've already noted, the action yields both a prediction and the confidence in that prediction. Once the model's accuracy reaches a level you're comfortable with, you can begin using the confidence score as an action condition, so that human tasks are only assigned if the confidence score is below a threshold defined by you.
Depending upon what type of fields you're using for the prediction, and what type of field you're predicting, I'm not sure how much variation you'll see after just a few additional rows are added to the dataset.
Thomas
1
Answers
-
@Jeff_146001, the prediction model was trained successfully, thanks for your hint.
0 -
@Thomas_937381 , thanks for your answer. understood
0