1 / 18

On Subgroup Discovery in Numerical Domains

On Subgroup Discovery in Numerical Domains. Henrik Grosskreutz and Stefan Rüping Fraunhofer IAIS. TexPoint fonts used in EMF. Read the TexPoint manual before you delete this box.: A A A A. Overview. Introduction: Subgroup discovery in numerical domains Motivation and definition

babu
Télécharger la présentation

On Subgroup Discovery in Numerical Domains

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. On Subgroup Discovery in Numerical Domains Henrik Grosskreutz and Stefan Rüping Fraunhofer IAIS TexPoint fonts used in EMF. Read the TexPoint manual before you delete this box.: AAAA

  2. Overview Introduction: Subgroup discovery in numerical domains Motivation and definition Problems New algorithm: Based on a new pruning scheme Empirical evaluation

  3. Subgroup Discovery – A Local Pattern Discovery Task Task: Find descriptions of subgroups of the overall population that are both Large Unusual label Subgroup description: Conjunction of attribute-value pairs Example: Profession= Teacher & Sex=M Cost=High Example: Employee dataset Class attribute

  4. The Quality of a Subgroup Quality Functions: na(p - p0) n: Size of the subgroup extension p: target share in subgroup p0: overall target share 0 a 1: Constant a=1: “Piatetsky-Shapiro”/”WRAC” a=0.5: Binomialtest Ex.: Profession= Teacher & Sex=M P-S Quality = 2 (1 – 4/6)

  5. Subgroup Discovery in Numerical (or Ordinal) Domains In many domains, the attributes are not nominal but numeric (or ordinal) Subgroup descriptions: Conjunctions of attribute-interval pairs Examples BMI [26,30] and BP [160,180] BMI [26,30] Example: Medical dataset

  6. How to deal with numerical attributes? Standard approach: (Entropy) Discretization (A) Replace every numeric attribute by one nominal attribute ranging over non-overlapping intervals Result D(X’) = {[1-2],[3-4],[5,6]} Problems Expected result should include • X  [1-4] and Y  [1-2] But we only obtain: • X = [1-2] and Y = [1-2] , • X = [3-4]

  7. How to deal with numerical attributes (ii)? Standard approach: Entropy Discretization (B) Replace every numeric attribute by a setofbinominal attributes, i.e. use overlapping intervals D(Y’) = {[1-2],[3-4],[5,6],[1-4],[1-6],[3-6]} D(X’) = Ø Problem: Expected solution should include: • X  [1-4] and Y  [1-2] But the obtained result will not contain any constraint on X

  8. Discretization Strategies on Benchmark Datasets: Quality of the Best Subgroup ‘diabetes‘ dataset ‘yeast ‘ dataset

  9. To find optimal subgroup descriptions, we have to consider attribute-interval constraints over arbitrary intervals How can we improve the performance if overlapping intervals are considered?

  10. Subgroup Discovery as Search in the Space of Subgroup Descriptions Empty description Ø Constraints based on X Constraints based on Y Subgroup descriptions of length 1 Can we make use of properties of the search space to speedup the computation? y  [y1,y2] y  [y1,ym] x  [x1,x2] x  [x2,x3] x  [x1,x3] x  [x1,xn] … y  [y1,y2] y  [y1,ym] y  [y1,y2] y  [y1,ym] Subgroup descriptions of length 2 Constraints based on Y,in conjunction with constrains on X y  [y1,y2] y  [y1,ym] y  [y1,y2] y  [y1,ym]

  11. Standard Approach: DFS with Optimistic Estimate Pruning (Horizontal Pruning) Empty description Ø Calculate OE(x  [x1,x3]) OE <   Prune branch Subgroup descriptions of length 1 y  [y1,y2] y  [y1,ym] x  [x1,x2] x  [x2,x3] x  [x1,x3] x  [x1,xn] … y  [y1,y2] y  [y1,ym] y  [y1,y2] y  [y1,ym] Subgroup descriptions of length 2 y  [y1,y2] y  [y1,ym] y  [y1,y2] y  [y1,ym]

  12. New Approach: “Horizontal Pruning” Calculate m1 = max. quality of refinements of x  [x1,x2] Calculate m2 = max. q. of ref. of x  [x2,x3] Empty description Ø Subgroup descriptions of length 1 y  [y1,y2] y  [y1,ym] x  [x1,x2] x  [x2,x3] x  [x1,x3] x  [x1,xn] … Use m1 and m2 to calculate tighter estimate OE(x  [x1,x3]) OE <   Prune branch y  [y1,y2] y  [y1,ym] y  [y1,y2] y  [y1,ym] Subgroup descriptions of length 2 y  [y1,y2] y  [y1,ym] y  [y1,y2] y  [y1,ym]

  13. Lemma Let tl and tr be two split points. The quality na(p - p0), (a1) of every refinement of sd  X  [tl, tr] is bound by the sum of the maximum of the qualities of all refinements of sd  X [tl, t] and sd  X [tl, t] for every tin [tl, tr] Ø This property • Also holds if a depth limitis considered • Also holds if an arbitrary set of candidate split points is considered X  [tl,tr] x  [tl,t] x  [t,tr] ≤ +

  14. Combining Exact Bounds and Classic Optimistic Estimates Empty description Ø Use maxQ below x  [x1,x2]and OE(x  [x2,x3])to prune x  [x1,x3] Subgroup descriptions of length 1 y  [y1,y2] y  [y1,ym] x  [x1,x2] x  [x2,x3] x  [x1,x3] x  [x1,xn] … y  [y1,y2] y  [y1,ym] y  [y1,y2] y  [y1,ym] Subgroup descriptions of length 2 y  [y1,y2] y  [y1,ym] y  [y1,y2] y  [y1,ym]

  15. A New Algorithm for SD in Numerical Domains Main Idea: DFS in space of subgroup descriptions Uses optimistic estimates to initialize the Bound tables When new bound maxQ for subgroups involving X  [tl,tr] becomes available Set Bound[l,r] to maxQ  l’, r’. l’ ≤ l, r ≤ r’ Bound [l’, r’] = Bound [l’, r]+maxQ+Bound [l, r’]. Heuristik: Never consider an interval if its subinterval have not yet been considered

  16. Experimental Results Number of nodes considered when using frequency discretization with 10 bins Pruning affects subgroup descriptions of length 1! diabetes transfusion

  17. Experimental Results (ii): Speedup mamography Speedup @ maximum length 2 transfusion

  18. Summary Classical subgroup discovery approaches… … have problems in numeric domains. New “horizontal” pruning scheme speeds up computation • Open Issues • Good Sets of Subgroups (Iterated Weighted Covering?) • Mixed Domains • Further Speedups • … Thank you for your attention!

More Related