Merge pull request #6 from christophM/cleanup

Rework API and prepare CRAN release

.Rbuildignore

@@ -2,6 +2,7 @@
 ^\.Rproj\.user$
 README.*
 README_files
+cran-comments.md
 framework.jpg
 .travis.yml
 Makefile

.gitignore

@@ -5,3 +5,6 @@
 *.swp
 iml.Rcheck
 ..Rcheck
+README.html
+notes.md
+

DESCRIPTION

@@ -1,15 +1,13 @@
 Package: iml
 Type: Package
-Title: Methods for interpretable machine learning
-Version: 0.1
-Date: 2017-11-30
-Author: Christoph Molnar
+Title: Interpretable Machine Learning
+Version: 0.2
+Date: 2018-03-04
 Authors@R: c(person(given = "Christoph", family = "Molnar", 
                     role = c("aut", "cre"), email = "[email protected]"))
 Maintainer: Christoph Molnar <[email protected]>
-Description: Model-agnostic interpretability methods for machine learning models. 
-  The iml package provides you with tools to analyse and explain the behaviour and prediction of black box machine learning models. 
-  Implemented tools are: partial dependence plots, local models (LIME), feature importance measure and many more. 
+Description:The iml package provides model-agnostic interpretability methods to
+ analyse and explain the behaviour and predictions of any black box machine learning model. 
 Imports: R6,
          checkmate,
          dplyr, 
@@ -18,16 +16,15 @@ Imports: R6,
          partykit,
          glmnet, 
          Metrics, 
-         data.table,
-         mlr
+         data.table
 Suggests: randomForest,
           gower, 
           testthat, 
           rpart, 
           MASS,
           caret,
           e1071, 
-          lime
+          lime,
+          mlr
 License: MIT + file LICENSE
-Roxygen: list(wrap = FALSE)
 RoxygenNote: 6.0.1

LICENSE

@@ -1,21 +1,2 @@
-MIT License
-
-Copyright (c) [year] [fullname]
-
-Permission is hereby granted, free of charge, to any person obtaining a copy
-of this software and associated documentation files (the "Software"), to deal
-in the Software without restriction, including without limitation the rights
-to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-copies of the Software, and to permit persons to whom the Software is
-furnished to do so, subject to the following conditions:
-
-The above copyright notice and this permission notice shall be included in all
-copies or substantial portions of the Software.
-
-THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE.
+YEAR: 2018
+COPYRIGHT HOLDER: Christoph Molnar

NAMESPACE

@@ -1,14 +1,15 @@
 # Generated by roxygen2: do not edit by hand
 
-S3method(plot,Importance)
-S3method(predict,LIME)
+S3method(plot,FeatureImp)
+S3method(predict,LocalModel)
 S3method(predict,TreeSurrogate)
-export(feature.imp)
-export(ice)
-export(lime)
-export(pdp)
-export(shapley)
-export(tree.surrogate)
+export(FeatureImp)
+export(Ice)
+export(LocalModel)
+export(PartialDependence)
+export(Predictor)
+export(Shapley)
+export(TreeSurrogate)
 import(Metrics)
 import(R6)
 import(checkmate)
@@ -21,9 +22,7 @@ importFrom(dplyr,left_join)
 importFrom(dplyr,one_of)
 importFrom(dplyr,summarise)
 importFrom(glmnet,glmnet)
-importFrom(mlr,getPredictionProbabilities)
-importFrom(mlr,getPredictionResponse)
-importFrom(mlr,getTaskType)
+importFrom(stats,model.matrix)
 importFrom(stats,predict)
 importFrom(tidyr,gather)
 importFrom(utils,methods)

NEWS.md

@@ -0,0 +1,11 @@
+# iml 0.2
+* The API has been  reworked: 
+  * User directly interacts with R6 classes (`pdp()` is now `PartialDependence$new()`).
+  * User has to wrap the machine learning model with `Predictor$new()`.
+  * New data points in `Shapley` and `LocalModel` can be set with `$explain()`.
+  * `Lime` has been renamed to `LocalModel`.
+* Plots have been improved.
+* Documentation has been improved.
+
+# iml 0.1
+Initial release

R/DataSampler.R → R/Data.R

@@ -1,4 +1,4 @@
-DataSampler  = R6::R6Class('DataSampler',
+Data  = R6::R6Class("Data",
   public = list(
     X = NULL,
     y = NULL,
@@ -8,36 +8,44 @@ DataSampler  = R6::R6Class('DataSampler',
     n.features = NULL,
     prob = NULL,
     sample = function(n=100, replace = TRUE, prob = NULL, get.y=FALSE) {
-      if(is.null(prob) & !is.null(self$prob)) {
+      if (is.null(prob) & !is.null(self$prob)) {
         prob = self$prob
       }
       indices = sample.int(private$nrows, size = n, 
         replace = replace, prob = prob)
-      if(get.y){
+      if (get.y) {
         cbind(self$X[indices,], self$y[indices,])
       } else {
         self$X[indices, ]
       }
     },
-    get.x = function(...){
+    get.x = function(...) {
       self$X
     },
-    get.xy = function(...){
+    get.xy = function(...) {
       cbind(self$X, self$y)
     },
-    print = function(){
+    print = function() {
       cat("Sampling from data.frame with", nrow(self$X), "rows and", ncol(self$X), "columns.")
     },
-    initialize = function(X, y = NULL, prob = NULL){
+    initialize = function(X, y = NULL, prob = NULL) {
       assertDataFrame(X, all.missing = FALSE)
-      assert_named(X)
-      assertDataFrame(y, all.missing = FALSE, null.ok = TRUE)
+      assertNamed(X)
       self$X = X
-      if(!missing(y)){
+      if (length(y) == 1 & is.character(y)) {
+        assert_true(y %in% names(X))
+        self$y = X[y]
+        self$y.names = y
+        self$X = self$X[setdiff(colnames(X), self$y.names)]
+      } else if (inherits(y, "data.frame")) {
+        assertDataFrame(y, all.missing = FALSE, null.ok = TRUE, nrows = nrow(X))
         self$y = y
-        self$y.names = names(y)
-        assert_true(nrow(X)==nrow(y))
-      }
+        self$y.names = colnames(self$y)
+      } else if (is.vector(y) | is.factor(y)) {
+        assert_vector(y, any.missing = FALSE, null.ok = TRUE, len = nrow(X))
+        self$y = data.frame(..y = y)
+        self$y.names = colnames(self$y)
+      } 
       self$prob = prob
       self$feature.types = get.feature.type(unlist(lapply(X, class)))
       self$feature.names = colnames(X)