(ns code.experiments.prep
  (:use (code.utils utils))
  (:use (code.learners.kmeans k_means_classifier k_means_model))
  (:use (code.data data standard_data))
  (:use (incanter core stats)))

(def numeric-datasets [(vector 4 paint3)
		       (vector 4 paint5)
		       (vector 4 paint10)
		       (vector 4 paint20)
		       (vector 4 paint37)
		     ;  (vector 4 iris)
		       (vector 4 wind204)])
		       
(def cat-datasets [(vector 4 audio)         ;0
                   (vector 4 bc)						;1
                   (vector 4 heart)					;2
                   (vector 4 lym)						;3
                   (vector 4 pima)					;4
                   (vector 4 bc-fss)				;5
                   (vector 4 heart-fss)			;6
                   (vector 4 lym-fss)				;7
                   (vector 4 pima-fss)			;8
                   (vector 4 lym5-fss)			;9
                   (vector 4 lym6-fss)			;10
                   (vector 4 lym8-fss)			;11
                   (vector 4 lym10-fss)			;12
                   (vector 4 tumor)					;11
                   (vector 4 splice)])			;12

(def ndatasets (map #(vector (first %) (make-random-data (second %))) numeric-datasets))
(def cdatasets (map #(vector (first %) (make-random-data (second %))) cat-datasets))
(def cat cat-euclidean-distance)
(def numeric numeric-euclidean-distance!)

(defn exp1 [k data distance]
  (let [clusters (k-means-model k (shuffle data) distance)]
    (loop [i 0 cluster clusters results (transient [])]
      (if (empty? cluster)
	(matrix (apply concat (persistent! results)))
	(recur 
	 (inc i)					
	 (rest cluster)
	 (conj! results (let [answer (bind-columns 
				      (first cluster) 
				      (matrix (repeat (nrow (first cluster)) i)))]
			  (if (= (nrow answer) 1) [answer] answer))))))))