In all these cases the algorithm terminated on the 1st time that it matched the pattern it was aiming for. There are many use-cases where this may not be the most desirable behaviour. For instance, a holiday company may want to select people who have been to Australia-United States-Greece in the last 12 months. If a person had this sequence 10 years ago, and also 10 months ago then it would be the first (oldest) of these patterns that would be matched – and the person would croatia phone number example not be selected. In this example, we would want to select based upon the Last match, and not the first match (*1).
In this blog we explain how we can go about doing this, and illustrate some of the analytical questions that we can now seek to answer.
Returning the first matched pattern is relatively straightforward. We can simply break off from scanning through the transactions as soon as we know we have encountered the pattern we are looking for. In some of the more generalised cases discussed below we have to scan through all of the transactions, find and store all the matched patterns and then sort them in to a specific order before extracting details of one of the matched patterns.
The user will get presented with the following choices in the Pattern Match controls. The bottom value is the pre-existing choice of the value from the matched pattern to return. The top 3 controls give the user the ability to specify which particular pattern they want to match.