AI-Assisted Data Reconciliation
Data reconciliation might sound trivial, just compare one data source to the other and see if they match. Unfortunately, same products in different systems (for example CRM and billing) are often called differently. In this case you need to have a list of matching rules, translating between the two. In order to generate rules candidates we use a leaner, but parallelized version of Apriori algorithm, which improves our data processing 10 times. To evaluate the generated rules, we use a Random Forrest algorithm, with similarity metrics as features. User gets a list of recommended rules and accompanying features that support the decision. A matching rule can then be confirmed with just a click.