Added k-nearest neighbor classifier (:lazy :ibk)

src/clj_ml/classifiers.clj

@@ -64,6 +64,7 @@
   (:import (java.util Date Random)
            (hr.irb.fastRandomForest FastRandomForest)
            (weka.core Instance Instances)
+           (weka.classifiers.lazy IBk)
            (weka.classifiers.trees J48 RandomForest M5P)
            (weka.classifiers.meta LogitBoost AdditiveRegression RotationForest)
            (weka.classifiers.bayes NaiveBayes NaiveBayesUpdateable)
@@ -78,6 +79,15 @@
   "Creates the right parameters for a classifier. Returns the parameters as a Clojure vector."
   (fn [kind algorithm map] [kind algorithm]))
 
+(defmethod make-classifier-options [:lazy :ibk]
+  ([kind algorithm m]
+     (->> (check-options m
+                         {:inverse-weighted "-I"
+                          :similarity-weighted "-F"
+                          :no-normalization "-N"})
+          (check-option-values m
+                               {:num-neighbors "-K"}))))
+
 (defmethod make-classifier-options [:decision-tree :c45]
   ([kind algorithm m]
      (->> (check-options m
@@ -285,6 +295,7 @@
 
    The classifiers currently supported are:
 
+     - :lazy :ibk
      - :decision-tree :c45
      - :decision-tree :boosted-stump
      - :decision-tree :boosted-decision-tree
@@ -304,6 +315,25 @@
    This is the description of the supported classifiers and the accepted
    option parameters for each of them:
 
+    * :lazy :ibk
+
+      K-nearest neighbor classification.
+
+      Parameters:
+
+        - :inverse-weighted
+            Neighbors will be weighted by the inverse of their distance when voting. (default equal weighting)
+            Sample value: true
+        - :similarity-weighted
+            Neighbors will be weighted by their similarity when voting. (default equal weighting)
+            Sample value: true
+        - :no-normalization
+            Turns off normalization.
+            Sample value: true
+        - :num-neighbors
+            Set the number of nearest neighbors to use in prediction (default 1)
+            Sample value: 3
+
     * :decision-tree :c45
 
       A classifier building a pruned or unpruned C 4.5 decision tree using
@@ -430,6 +460,10 @@
 "
   (fn [kind algorithm & options] [kind algorithm]))
 
+(defmethod make-classifier [:lazy :ibk]
+  ([kind algorithm & options]
+     (make-classifier-with kind algorithm IBk options)))
+
 (defmethod make-classifier [:decision-tree :c45]
   ([kind algorithm & options]
      (make-classifier-with kind algorithm J48 options)))

test/clj_ml/classifiers_test.clj

@@ -2,6 +2,24 @@
   (:use [clj-ml classifiers data] :reload-all)
   (:use clojure.test midje.sweet))
 
+(deftest make-classifiers-options-ibk
+  (fact
+   (let [options (make-classifier-options
+                  :lazy :ibk
+                  {:inverse-weighted true :similarity-weighted true :no-normalization true :num-neighbors 3})]
+     options => (just ["-I" "-F" "-N" "-K" "3"] :in-any-order))))
+
+(deftest make-classifier-ibk
+  (let [c (make-classifier :lazy :ibk)]
+    (is (= (class c)
+           weka.classifiers.lazy.IBk))))
+
+(deftest train-classifier-ibk
+  (let [c (make-classifier :lazy :ibk)
+        ds (clj-ml.data/make-dataset "test" [:a :b {:c [:m :n]}] [[1 2 :m] [4 5 :m]])]
+    (clj-ml.data/dataset-set-class ds 2)
+    (classifier-train c ds)
+    (is true)))
 
 (deftest make-classifiers-options-c45
   (fact