https://forum.knime.com/t/pattern-analysis/14654/4 Hi guys, I have a question regarding the approach that was mentioned in the workflow shared in the link above. The part where the columns where combined is really helpful, but I have troubles regarding the positive. @mlauber71 mentioned some ways to do so, but could you help me out by doing this without a positive list, because my dataset is to big so therefore it is impossible to create a positive list manually. Which nodes can I use to take all patterns which were combined as a positive list. But it should only take the patterns which occurs more than three times.i think i can find out the number of individual patterns with the help of node Extract table Dimension, or? after that I can use the java edit variable node to filter out the patterns where the number of occurrence is under 3. After that I can use the group by node to bring the patterns together to have just unique patterns. Does this make sense? One more question regarding the prediction if it is a repeating business or not: When the occurrence of a transaction (pattern) is more than three times, it is a repeating business, if not than it is not a repeating business. this would be the classification. But how can I use these created patterns to build a predictive model, I mean how can I connect it?. Thank you so much in advance. Kind regards, Canan

Correct, then the count is updated to include the strings added to the cluster and the clustering is repeated. Clustering is something you’ll do when you have no list of “correct” values to match to. In your Country names example here you may want to play around with the similarity search node using a Levenshtein Distance (it’s for strings). One other general note that I don’t think we addressed in the thread, it’s probably a good choice to lower or uppercase all your strings before getting to into the processing, a =/= A unfortunately! Best of luck!

pattern analysis

KNIME Analytics Platform

izaychik63 April 23, 2019, 1:29pm 4

Look here. It ca be very close to what you need.