Finding a list which has the most similar entries to a reference list

[BFernandez]

Hi Guys,

Found this a bit hard to explain, so any questions please don't hesitate to ask!

I have two lists of product data, one for buyers and one for sellers (these are listed as A, B, C).
The product names are not exactly the same (Eg Playstation and playstation three should be matched), I would have thought using the FuzzyLogic add in to match these would be the way forward but i'm open to suggests!
I need to rank the sellers by how many of their items appear on the buyers list

If anyone has a solution for this, but only finds exact matches I'd really like to hear it, I think I might be able to change the data so that the items are exactly the same

Thanks!

[newdoverman]

This does some of what you want but there are just too many possibilities for the economics entries. This compares column D against column B and if the contents of column D is found in column B it is counted. The entries for economics show the limitations of this.

Please Login or Register  to view this content.

[BFernandez]

Thanks Newdoverman, however not quiet there, apologies the explanation was poor, I need the match buy and seller lists which are most similar

For example, seller A has two matches with buy A; a bean bag and an international economics book, zero matches with seller B, and one match with seller C, a bean bag
Ergo, for seller A buyer A would be the closest match, followed by buyer C

[newdoverman]

There might be a solution to this problem but there are just too many possibilities of matches due to the need of using wildcards to make matches. A control over what is entered for the buyers' purchases would be the easiest to implement. Due to this, there really isn't any way of knowing who bought what from whom.

[BFernandez]

I can ensure that all similar items are the same if you know of a solution to matching these similar lists?

[newdoverman]

I suspect that this is a very small portion of a much larger listing and if that is the case, the problems of matching will grow exponentially. There has to be a way of limiting what is entered (within reason).

[BFernandez]

Yes the list could grow to a huge number of variations, but I'm working on a solution for that separately using a combination of FuzzyLogic and macro's, this can be done prior to the matching of a list so that it only needs to match exact strings.

Any suggestions from anyone?

Thanks,

[newdoverman]

With the first list in column A and the second list in column D

Enter in column B and copy down. I have used 20,000 rows but you can change that to suit.

Please Login or Register  to view this content.

Enter in Column C and copy down to check the entries in column D match the entries in column A.

Please Login or Register  to view this content.

Now you can sort to separate what is included in both lists and what isn't.

[BFernandez]

Thanks for the suggestion, but I think youve misunderstood slightly. Using the file attached as an example, Column A lists seller names (A, B, C are used in the example), column B lists the item each of the sellers items. For example, Seller A is selling a bean bag, and a International economics [textbook].
The same applies to column C and D respectively, the difference being that these are buying these products.

I am looking for a way to identify which buyer has the most similar items for each of the sellers. For example, seller A would have buyer A listened next to their items as their list contains all of their items