Changes to the Large Itemset Phase Sample Clauses

Changes to the Large Itemset Phase. In the original approach, Large itemsets were retrieved in this phase, however, since con- secutiveness is required for our cause, the search for crimesets that contain more than one crime could already be seen as a search for subcareers. For example, if the crimeset 12 is Large, this implies that crime 2 immediately follows crime 1. Since the search for se- quentiality only starts in phase 4, we should limit our approach to the search for singular items in this case. An exception to this rule is the situation that both 12 and 21 are Large, indicating that the crimeset 12 is Large as well. However, the inclusion of this set provides no additional gain over the inclusion of separate crimesets 1 and 2; since the inclusion of 12 implies that all its subsets are also frequent per the AprIorI-property, we cannot delete extra individuals in the next phase based upon the discovery of a frequent 12. We can therefore safely limit our search for Large itemsets to the search for frequent crimes in the database. Another advantage to this situation is that there is no need to do any AprIorI search within the database since a simple count of each crime will yield the required results, greatly reducing the computational complexity of our approach. Another difficulty that is overcome by the search for single Large crimes alone, is the confusion about the meaning of the statement “12 is Large”. This could either indicate that the summation of occurrences of 12 and 21 is above the threshold, or that both of their occurrences are above the threshold, the first representing a more loose statement and the latter describing a more strict situation. We therefore restrict our search to the retrieval of singular crimes in this phase or simple consecutive subcareers (without overlining) in the next phase.