;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
; 
; Assignment 6 Problem 3
;
; Here's some information on how I solved this problem (in order to
; see how well you could do with this problem.)  BTW, my code for this
; problem discretizes all 30 attributes using a single threshold for
; each attribute.  I wrote it in 80 lines (including comments) in
; about an hour.
;
;


; I wrote a the following function which automatically generates a
; discretize procedure.  I'm not giving you all the other procedures
; that this one calls, but you can get the general idea of my approach
; from this.
;
;
(define (create-discretize-fn tdata)
  ; find thresholds for n elements (0 through n-1) and return a list of them
  (define (find-thresholds i n)
    (if (= i n)
	'()
	(cons (find-best-threshold-val tdata i)
	      (find-thresholds (+ i 1) n))))
 
  (let ((thresh-list (find-thresholds 0 (length (second (first tdata))))))
    (display "Here is the list of threshold values:\n")
    (display thresh-list) ; print out the list so I can see it
    ; here's my discretize procedure
    (lambda (ex)
      (map (lambda (a b) 
	     (if (>= a b)
		 'high
		 'low))
	   ex
	   thresh-list))))


; one of the procedures I wrote picks the best threshold value using
; the procedure we discussed in class.
;
(find-best-threshold-val bc-data-s 9)
;Value: .05915


; Now I'll run my create-discretize-fn on two of the training data sets...
;
(define discretize-fn (create-discretize-fn bc-data-l1))
Here is the list of threshold values:
; I deleted the values that were printed here so I'm not just giving you
; the answer!

;Value: discretize-fn

; The function that would go in my file to turn in for problem 3 would
; simply take the values printed out above and hardcode them into my
; discretize function.  
(define (discretize attribute-values)
  (let ((thresh-list ; I inserted values from above here!))
    (map (lambda (a b) 
	   (if (>= a b)
	       'high
	       'low))
	   attribute-values
	   thresh-list)))


; learn a decision tree using the discretize procedure...
(define dt (ldt-disc discretize bc-data-l1))
;Value: dtr1

; if you've got the latest version of the a6code, you can turn of all
; the printing that would normally appear by setting the following variable
(define print-info #f)
;Value: print-info

; now test the decision tree on another data set
(testdt-disc discretize dt bc-data-l3)
;Value 26: (194 200)

; it correctly classified 194 out of 200 examples...