Upload historical data for purchase protection
In Microsoft Dynamics 365 Fraud Protection's evaluate and protect experiences, you can upload your purchase protection historical data into the system to help train and improve the accuracy of the machine learning models. The uploads can include data for purchases, chargebacks, merchant and bank decisions, and accounts.
For the best results, we recommend that you upload at least six months' worth of historical data.
You can upload historical data through the Fraud Protection portal or the application programming interface (API).
- Payment instruments
- Update accounts
- Update account addresses
- Update account payment instruments
This data is sensitive, and you should take care to upload it only from a secure network location. Be aware that Microsoft requests only partial data about payment instruments (the bank identification number [BIN] and the last four digits). We don't request the full payment instrument number or Social Security number (SSN). We recommend that you not include this type of data in the files that you upload. For more information about how data is used and protected in Fraud Protection, see Security, compliance, and data subject requests.
You upload your historical data on the Purchase protection tab of the Data upload page.
To ensure that Fraud Protection correctly interprets your files, follow these requirements, and review the required schemas:
- The files are in CSV UTF-8 (comma-delimited) format (*.csv).
- The maximum file size is 10 gigabytes (GB).
- DateTime columns are in International Organization for Standardization (ISO) 8601 format.
- The decimal precision is two decimal places.
- The following characters are escaped in all columns: commas, new line characters, and multiline characters.
To browse and upload data files:
In the left navigation, select Data, and then select Data upload.
On the Data upload page, on the Purchase protection tab, select Select data source.
Select the file format type: CSV or TSV.
If the file has no column headers, select the This file doesn't include column headers option.
The system will automatically assign default column headers.
To begin data mapping, select Next, and then select the columns in your file that you want to map to each schema attribute.
In the data preview pane at the bottom of the page, confirm that the mapping is correct.
When all the attributes are mapped, select Save and close to return to the Data upload page.
Select Process data.
You can save an incomplete mapping and then return to the Purchase protection tab later to complete it.
For related data types, such as purchase and account data, every file must be uploaded before you process the data.
To upload additional data files:
- On the Data upload page, on the Purchase protection tab, select Edit.
- Select Change data source to upload more files.
- To apply the currently saved mapping to the newly uploaded file, select Keep the current mapping.
- After mapping is completed, select Save and close.
- On the Data upload page, select Process data.
Data processing might take up to a few hours, depending on the size of the file. You can check the status of data processing for each data source. If data processing encounters data errors, processing will fail, and the data import will be incomplete. Review the error file details, and fix the data issue that is noted. After you've finished fixing the issues, re-upload the file, remap the attributes if remapping is required, and retry data processing.
To remove previously uploaded data files:
- On the Data upload page, on the Purchase protection tab, select the files to remove, and then select Delete.
During the evaluate and protect experiences, data can also be ingested through the API. A score will be returned for data that you upload in this way. For a more comprehensive overview of this process, see Integrate Dynamics 365 Fraud Protection APIs.
Download sample data
Before you use your internal data, you can download sample data and use it to explore Fraud Protection options.
To download sample data: