FAQs for Intelligent Account Combination Defaulting
When does the Intelligent Account Combination Defaulting feature predict values?
For the feature to predict segment values, you need to complete all the steps required to activate it. See Activate the Feature section in Intelligent Account Combination Defaulting.
How do I know that the feature is predicting values?
The feature predicts values when you click or navigate to the distribution combination fields at the invoice line level. The prediction service has a time-out of 2 seconds. Usually, the segment values are defaulted in less than 2 seconds. After the default segment values are predicted, the distribution combination fields are highlighted for your review. If you don't see predictions within 2 seconds, you can continue entering values in the fields.
What's the prediction coverage and accuracy rate of the feature?
The number of times a segment defaults with predictions and the accuracy of the predictions depend on the invoice data and the segment. Some segments like Account and Cost Center are more difficult to predict and might have a lower prediction coverage and accuracy. On an average, you can expect 60% to 70% prediction coverage and 97% to 98% accuracy rate in the predictions.
How can I improve prediction coverage and accuracy?
The machine learning models make better predictions when the input data is consistent. The models use invoice header and line information such as the supplier name, supplier site, amounts, and descriptions to predict different segment values. Prediction accuracy is better with patterns of consistency. Using Oracle Intelligent Document Recognition, IDR, helps in better data consistency which helps in better prediction results.
What are the things I need to check in my invoices to increase the number of defaulted segments?
- Enter all invoice information at the header and line level. You need not capture additional data but enter all the usual data attributes. Predictions done with fewer data attributes will impact the confidence of the prediction and fewer segments are defaulted.
- In your invoice, if the description and other fields are after the Distribution Combination field, it's best practice to reorder the invoice layout and move the description and other fields to appear before the Distribution Combination field. By default, the line description is displayed after the Distribution Combination field.
What data does the machine learning model use for providing predictions using Intelligent Account Combination Defaulting?
The Intelligent Account Combination Defaulting feature leverages multiple input attributes from the invoice header and invoice line level to develop an efficient algorithm. These attributes include Supplier, Supplier Site, Business Unit, Legal Entity, Amounts, Currency, Descriptions, Requester and Invoice Source, and Line Source.
The list of attributes that the model uses might change over time as the algorithm is enhanced. The algorithm decides which attributes are the most relevant to make a prediction for each segment and uses them. Consistency in data patterns helps the algorithm enhance faster and provide better predictions.
Does the Intelligent Account Combination Defaulting feature support the Invoice Coding Workflow?
Intelligent Account Combination Defaulting currently doesn’t support the Invoice Coding Workflow. These code combinations are entered at the invoice distribution level and the predictions are only made at the invoice line level.
If the invoices are routed for coding which uses the invoice approval workflow, the coding is performed at the invoice line level and the feature predicts segment values.
Can I change the values predicted and defaulted by the feature?
Yes, you can change the values defaulted by the feature. If you delete the predicted value and enter a different value, the feature doesn’t overwrite it. Instead, the model uses it as an input to enhance the algorithm.
If you want the feature to provide predictions after you overwrite the defaulted values, delete all the segment values, click anywhere else on the screen, and then navigate back to the distribution combination field.
I corrected the code combination that was predicted and defaulted by the feature after a few weeks. Will the model take the new values as an input?
Yes, the machine learning model algorithm takes the corrected values as an input and is trained on the latest valid data recorded in Payables.
What are the best practices to test the Intelligent Account Combination Defaulting feature?
In a test environment, we recommend that you validate the feature on data that's similar to production data since the algorithms are trained on data that’s copied from production. In the test environment, the algorithms don’t continue to be trained on new data that's entered after the P2T Refresh date and hence the accuracy or coverage won't change significantly based on the new transactions that you entered.
When you implement the feature in a test environment, the best way is to use test invoices that use the same invoice attributes as production invoices. It’s best practice to not use test data such as 1 USD invoice value and a description value such as Test.
If the Chart of Accounts definitions change, how long does the model take to learn the new definitions and predict values?
If you change a default segment value, the algorithm recognizes and starts to modify the model, but it needs more examples to predict default values based on the changed definitions. The algorithm defaults a value when the confidence score of the prediction crosses the threshold value for that segment. Each change you make to a defaulted value modifies the confidence score of the model. After the model receives enough examples and the confidence score crosses the threshold value, the new value will be defaulted.
The time that the model needs to learn the new pattern depends on the number of invoices and the consistency of data. High volumes of invoices with consistent data can result in accurate predictions within days.
Why can’t I access the My Enterprise section in Oracle Financials Cloud?
-
ASM_REVIEW_APPLICATIONS_OFFERINGS_PRIV
-
ASM_CONFIGURE_OFFERING_PRIV
After you add these privileges to the Implementation Consultant role, you can access the My Enterprise section and complete the activation of the feature.