Day 15_Model_Evaluation_diabetesDataset

# -*- coding: utf-8 -*-
"""Day 21_Diabetes_MultiClass.ipynb

Automatically generated by Colaboratory.

Original file is located at
    https://colab.research.google.com/drive/1HeSxItiMdChLlBldpyQI9T8whLYokEPd
"""
 
from google.colab import drive
drive.mount('/gdrive')
 
PATH = "/gdrive/My Drive/Colab Notebooks/resources/"
 
import matplotlib.pyplot as plt
from tensorflow import keras
import tensorflow as tf
import numpy as np
 
print(tf.__version__)
 
xdata = np.transpose([[1,2],
         [2,3],
         [3,1],
         [4,3],
         [5,3],
         [6,2]])
 
ydata = np.transpose([ [0],[0],[0],[1],[1],[1] ]) 
 
X = tf.placeholder(tf.float32, shape = [ 2, None ])
y = tf.placeholder(tf.float32, shape = [ 1, None ])
 
W1 = tf.Variable(tf.random.normal([1, 2]))
B1 = tf.Variable(tf.random.normal([1]))
 
Z1 = W1 @ X + B1
A1 = tf.sigmoid(Z1)
 
loss = -tf.reduce_mean( y * tf.log( A1 ) + (1 - y) * (tf.log(1 - A1) ) ) 
 
# dW1  = tf.gradients(loss, W1)[0]
# dB1  = tf.gradients(loss, B1)[0]
 
# lr = 0.01
# update_W1 = tf.assign(W1 , dW1 - lr * dW1) 
# update_B1 = tf.assign(B1 , dB1 - lr * dB1) 
 
 
train = tf.train.GradientDescentOptimizer(0.01).minimize(loss)
 
# 첫번째 인수를 두번째 인수로 캐스팅하라.
predicted = tf.cast( A1 > 0.5 , dtype = tf.float32 ) 
accuracy  = tf.reduce_mean( tf.cast( tf.equal(predicted, y) , dtype = tf.float32 ) ) 
 
with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())
 
    for step in range(10001):
        _ , cv = sess.run([train, loss], feed_dict = { X : xdata, y : ydata })
        # _ , cv, _, _ = sess.run([train, loss, update_W1, update_B1 ], feed_dict = { X : xdata, y : ydata })
        
        if step % 200 == 0:
            print(cv)
 
        hv, pv, av = sess.run([ A1, predicted, accuracy ], feed_dict = {X : xdata, y : ydata})
        print ("hf = " , hv, "pred = ", pv, "acc : ", av)
 
df = np.loadtxt(PATH + "data/diabetes.csv", delimiter= ",")
 
xdata  = df[:,:-1].T
ydata  = df[:, -1]
 
xdata.shape, ydata.shape
 
n_h = 1
n_x = 8
n_y = 1
 
X  = tf.placeholder(tf.float32, shape = [ n_x, None ] )
y  = tf.placeholder(tf.float32, shape = [  None ] )
W1 = tf.Variable(tf.random.normal([ 1, n_x ]) )
B1 = tf.Variable(tf.random.normal( [ 1 ] ) )
 
Z1 = W1 @ X + B1
A1 = tf.sigmoid( Z1 )
 
cost  = -tf.reduce_mean( y * tf.log( A1 ) + (1 - y) * tf.log( 1- A1 )    )
train = tf.train.GradientDescentOptimizer(0.01).minimize( cost )
 
predicted = tf.cast( A1 > 0.5, dtype = tf.float32 ) 
accuracy  = tf.reduce_mean(tf.cast(tf.equal(predicted ,y)  , dtype = tf.float32))
 
with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())
 
    for step in range(20001):
        if step % 2000 == 0:
            cv, _ = sess.run([cost, train], feed_dict = { X : xdata, y : ydata})
            print(step, cv)
    hv, pv, av = sess.run([A1, predicted, accuracy],feed_dict = { X : xdata, y : ydata })
    print("hv : ", hv, "pv : ", pv , "acc : ", av)
 
m = xdata.shape[1]
indices = np.random.permutation(m)
k = 3
 
err = 0
for i in range(k):
 
    test_indices  = indices[ int( i * ( m/k )) : int((i + 1) * (m / k) - 1)]
    train_indices = np.setdiff1d( indices, test_indices )
 
    with tf.Session() as sess:
        sess.run(tf.global_variables_initializer())
 
        for step in range(20001):
            if step % 2000 == 0:
                cv, _ = sess.run([ cost, train ], feed_dict = { X : xdata[:,train_indices], y : ydata[train_indices] } )
                # print(step, cv)
 
                
        preds = sess.run(accuracy,feed_dict = { X : xdata[:,test_indices], y : ydata[test_indices] })  
        print(preds)
        err += preds
 
print('total error : ', err/k)
 
"""# 시그모이드
시그마 X 에서 X 가 0 보다 작으면 결과는 0에 가까워지고
0보다 크면 y는 1에 가까워 진다.

시그모이드로 분류를 한다는 것은 y^ 이 y가 되는 결정경계를 찾는 과정이다.
"""