Announce: hlcm 0.2.2 - Parallel closed frequent itemsets mining

Dear all, I'm pleased to announce the release of hlcm on Hackage : http://hackage.haskell.org/package/hlcm-0.2.2 hlcm is data mining tool for computing closed frequent itemsets. This problem is famous as "market basket analysis": - given a list of transactions : [["bread", "butter","chocolate","tomato"] ,["bread","butter"] ,["bread","pencil","butter","chocolate"] ,["bread","butter","book"]] - and a minimal frequency threshold in [1..4], let's say 2: we want items that are sold together in at least 2 transactions hlcm will tell you that ["bread","butter","chocolate"] appears in 3 transactions and that ["bread","butter"] appears in 4 transactions. You can many funnier applications with your own data, for example log analysis, mining words in web pages, etc. You can see details on getting started with the program here: http://membres-liglab.imag.fr/termier/HLCM/hlcm/hlcm/Main.html The library documentation is here : http://membres-liglab.imag.fr/termier/HLCM/hlcm/HLCM.html hlcm is based on the most efficient algorithm for closed frequent itemset mining, LCM, which is much, much faster than the well-known Apriori algorithm (more details when following the pointers from the homepage: http://membres-liglab.imag.fr/termier/HLCM/hlcm.html). hlcm can also exploit parallelism through Strategies, with promising speedups. We still have more work to do in order to beat existing C/C++ implementations, but you can have a look at the paper that we submitted at Haskell Symposium this year for a detailed experimental study: http://membres-liglab.imag.fr/termier/HLCM/hlcm.pdf Don't miss out the section about the influence of RTS parameters on parallel performance. Feel free to send me an e-mail if you have any question about hlcm. Alexandre -- _____________________________________________________________ Alexandre Termier LIG (Laboratoire d'Informatique de Grenoble) Université Joseph Fourier 681 rue de la Passerelle B.P. 72, 38402 Saint Martin d'Hères (FRANCE) Phone: +33 4 76 82 72 07 Fax: +33 4 76 82 72 87 http://membres-liglab.imag.fr/termier/ _____________________________________________________________
participants (1)
-
Alexandre Termier