> On Jan 12, 2017, at 5:37 PM, Lorenzo Isella <lorenzo.ise...@gmail.com> wrote: > > Dear All, > I am fine tuning a Cubist model (see > https://cran.r-project.org/web/packages/Cubist/index.html). > I am a bit puzzled by its output. On a dataset which contains 275 > cases, I get non mutually exclusive rules. > E.g., in the output below, rules 2 and 3 cover all the 275 cases of > the data set and rule 1 overlaps partially. > Am I misunderstanding something?
It is doing the right thing. The rules are first derived from a regression tree and, in the process of pruning the rules, they can produce overlapping sets. When the rules overlap, the regression output is average across the active rules. Thanks, Max > Many thanks > > Lorenzo > > > > > Cubist [Release 2.07 GPL Edition] Thu Jan 12 23:10:40 2017 > --------------------------------- > > Target attribute `outcome' > > Read 275 cases (21 attributes) from undefined.data > > Model: > > Rule 1: [204 cases, mean 0.5393324, range 0 to 2.285714, est err > 0.2598495] > > if > home_copub_after_all <= 0.7142857 > host_copub_after_all <= 1.833333 > then > outcome = 0.1666667 + 0.9 home_copub_after_all > + 0.11 home_copub_before_all > > Rule 2: [259 cases, mean 0.7445303, range 0 to 3.166667, est err > 0.1866440] > > if > host_copub_after_all <= 1.833333 > then > outcome = 0.0433333 + 0.75 home_copub_after_all > + 0.33 host_copub_after_all + 0.37 > top_10_after_all > > Rule 3: [16 cases, mean 4.4285712, range 2.142857 to 8.857142, est > err 1.0346190] > > if > host_copub_after_all > 1.833333 > then > outcome = 1.595 + 1.03 top_10_after_all + 0.45 > home_copub_after_all > > > Evaluation on training data (275 cases): > > Average |error| 0.2678023 > Relative |error| 0.38 > Correlation coefficient 0.94 > > > Attribute usage: > Conds Model > > 100% 54% host_copub_after_all > 43% 100% > home_copub_after_all > 57% > top_10_after_all > 43% > home_copub_before_all > > > Time: 0.0 secs >> ______________________________________________ R-help@r-project.org mailing list -- To UNSUBSCRIBE and more, see https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.