meta (Q6033301)

**Author**: \N**Source**: Unknown - Date unknown \N**Please cite**: \N\N1. Title: meta-data\N\N2. Sources:\N(a) Creator:\NLIACC - University of Porto\NR.Campo Alegre 823\N4150 PORTO\N(b) Donor: P.B.Brazdil or J.Gama Tel.: +351 600 1672\NLIACC, University of Porto Fax.: +351 600 3654\NRua Campo Alegre 823 Email: statlog-adm@ncc.up.pt\N4150 Porto, Portugal\N(c) Date: March, 1996\N\N(d) Acknowlegements:\NLIACC wishes to thank Commission of European Communities\Nfor their support. Also, we wish to thank the following partners\Nfor providing the individual test results:\N\N- Dept. of Statistics, University of Strathclyde, Glasgow, UK\N- Dept. of Statistics, University of Leeds, UK\N- Aston University, Birmingham, UK\N- Forschungszentrum Ulm, Daimler-Benz AG, Germany\N- Brainware GmbH, Berlin, Germany\N- Frauenhofer Gesellschaft IITB-EPO, Berlin, Germany\N- Institut fuer Kybernetik, Bochum, Germany\N- ISoft, Gif sur Yvette, France\N- Dept. of CS and AI, University of Granada, Spain\N\N\N3. Past Usage:\N\N\NMeta-Data was used in order to give advice about which classification\Nmethod is appropriate for a particular dataset.\NThis work is described in:\N\N-"Machine Learning, Neural and Statistical Learning"\NEds. D.Michie,D.J.Spiegelhalter and C.Taylor\NEllis Horwood-1994\N\N- "Characterizing the Applicability of\NClassification Algorithms Using Meta-Level Learning",\NP. Brazdil, J.Gama and B.Henery:\Nin Proc. of Machine Learning - ECML-94,\Ned. F.Bergadano and L.de Raedt,LNAI Vol.784 Springer-Verlag.\N\N-"Characterization of Classification Algorithms"\NJ.Gama, P.Brazdil\Nin Proc. of EPIA 95, LNAI Vol.990\NSpringer-Verlag, 1995\N\N\N4. Relevant Information:n\NThis DataSet is about the results of Statlog project.\NThe project performed a comparative study between Statistical, Neural\Nand Symbolic learning algorithms.\N\NProject StatLog (Esprit Project 5170) was concerned with comparative\Nstudies of different machine learning, neural and statistical\Nclassification algorithms. About 20 different algorithms were\Nevaluated on more than 20 different datasets. The tests carried out\Nunder project produced many interesting results.\N\NAlgorithms DataSets\N------------------------- --------------------------\NC4.5 NewId Credit_Austr Belgian\NAC2 CART Chromosome Credit_Man\NIndCART Cal5 CUT DNA\NCN2 ITRule Diabetes Digits44\NDiscrim QuaDisc Credit_German Faults\NLogDisc ALLOC80 Head Heart\NkNN SMART KLDigits Letters\NBayesTree CASTLE New_Belgian Sat_Image\NDIPLO92 RBF Segment Shuttle\NLVQ Backprop Technical TseTse\NKohonen Vehicle\N\N\NThe results of these tests are comprehensively described in a book\N(D.Michie et.al, 1994).\N\N5. Number of Instances: 528\N\N6. Number of Attributes: 22 (including an Id#) plus the class attribute\N-- all but two attributes are continuously valued\N\N7. Attribute Information:\N1. DS_Name categorical Name of DataSet\N2. T continuous Number of examples in test set\N3. N continuous Number of examples\N4. p continuous Number of attributes\N5. k continuous Number of classes\N6. Bin continuous Number of binary Attributes\N7. Cost continuous Cost (1=yes,0=no)\N8. SDratio continuous Standard deviation ratio\N9. correl continuous Mean correlation between attributes\N10. cancor1 continuous First canonical correlation\N11. cancor2 continuous Second canonical correlation\N12. fract1 continuous First eigenvalue\N13. fract2 continuous Second eigenvalue\N14. skewness continuous Mean of |E(X-Mean)|^3/STD^3\N15. kurtosis continuous Mean of |E(X-Mean)|^4/STD^4\N16. Hc continuous Mean entropy of attributes\N17. Hx continuous Entropy of classes\N18. MCx continuous Mean mutual entropy of class and attributes\N19. EnAtr continuous Equivalent number of attributes\N20. NSRatio continuous Noise-signal ratio\N21. Alg_Name categorical Name of Algorithm\N22. Norm_error continuous Normalized Error (continuous class)\N\N\N8. Missing Attribute Values:\N\NNote that fract2 and cancor2 only apply to datasets with more than\N2 classes. When they appear as '?' this means a don't care value.\N\NSummary Statistics:\N\NAttribute Min Max Mean Std\NT 270 20000 4569.05 5704.01\NN 270 58000 10734.2 14568.8\Np 6 180 29.5455 36.8533\Nk 2 91 9.72727 19.3568\NBin 0 43 3.18182 9.29227\NCost 0 1 0.13636 0.35125\NSdRatio 1.0273 4.0014 1.4791 0.65827\NCorrel 0.0456 0.751 0.23684 0.1861\NCancor1 0.5044 0.9884 0.79484 0.15639\NCancor2 0.1057 0.9623 0.74106 0.269\NFract1 0.1505 1 0.70067 0.3454\NFract2 0.2807 1 0.70004 0.29405\NSkewness 0.1802 6.7156 1.78422 1.79022\NKurtosis 0.9866 160.311 22.6672 41.8496\NHc 0.2893 4.8787 1.87158 1.44665\NHx 0.3672 6.5452 3.34502 1.80383\NMcx 0.0187 1.3149 0.31681 0.33548\NEnAtr 1.56006 160.644 20.6641 35.6614\NNsRatio 1.02314 159.644 28.873 37.925