Training Match Predictions with Transaction Matching Assistance
Service Administrators use a Train Match Prediction job to train a prediction model that can predict potential matches for unmatched transactions in a match type.
During training, historical manual match data for the specified match type is analysed and a prediction model is created. Match process associated with the match type are identified and training is performed for each match process within the match type. If a match process has less than 2500 historical manual match 1 to 1 transactions, training proceeds to the next match process in the match type. After the training completes successfully, the trained model can be applied to new data for predicting potential matches.
It is recommended that you retrain the model periodically. Training can be on a monthly basis or when ever there are significant changes in match data. Regular training ensures that the model remains up-to-date and adapts to any trends or changes in the data, thereby leading to more accurate and reliable predictions. Monitor the volume of new matches and the quality of prediction results to determine the optimal frequency for retraining the prediction model.
Steps to Train a Match Prediction Model for a Match Type
- From the Home page, select Applications, and then Jobs.
- Select the Transaction Matching tab.
- From Actions, select Train Match
Prediction.
The Train Match Prediction dialog appears.
- In Match Type, select the match type that must be
trained.
Only approved match types are displayed in the drop-down list.
- Click Submit.
A Train Match Prediction job is created and you can view its progress.
A log file is generated after the training job completes. Click Complete to display the log file and review the results of the job. The log file shows the training status for each match process within the selected match type
The following log file shows that two out of three match processes in the match type were trained.
