Day 15_Diabetes_House_XGBoost

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

Automatically generated by Colaboratory.

Original file is located at
    https://colab.research.google.com/drive/1Lvl2OmvVarGZ3grmeeW-xFhHE7H8KmCn
"""
 
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
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
 
print(tf.__version__)
 
df = pd.read_csv("https://archive.ics.uci.edu/ml/machine-learning-databases/breast-cancer-wisconsin/wdbc.data",  header = None)
 
from sklearn import preprocessing 
 
label_encoder = preprocessing.LabelEncoder() 
df2 = df.copy()
 
## 문자를 숫자로 변경 B = 0 , M = 1
df2[1] = label_encoder.fit_transform(df[1])
 
df2
 
fig = plt.gcf()
fig.set_size_inches(10 * 2, 10 * 2)
 
sns.heatmap(df2.corr(),  annot = True)
 
## 상관관계 높은 순서대로 배열후 상위 10개의 피처 인덱스만 추출
top_10_featureIdx = np.argsort(df2.corr()[1])[::-1][:10]
 
## 상관관계 높은 순서대로 배열후 상위 10개의 피처 인덱스만 추출
# top_20_featureIdx = np.argsort(df2.corr()[1])[::-1].reset_index(drop = True)[:20]
 
df3 = df2.iloc[:,top_10_featureIdx] 
df3.shape
 
## 열 이름 초기화 0번이 타겟
df3.columns = [ i for i in range(df3.shape[1])]
df3.shape
 
## 정규화 및 전치
from sklearn.preprocessing import minmax_scale, StandardScaler, RobustScaler
 
# xdata = minmax_scale( df3.iloc[:,1 : ] )  #결과 : 0.68xxx
# xdata = RobustScaler().fit_transform(df3.iloc[:,1 : ]) #결과 : 0.63xxx
xdata = StandardScaler().fit_transform(df3.iloc[:,1 : ])
ydata = df3.iloc[:,0]
 
xdata = xdata.T
ydata = np.array(ydata).reshape((1, -1))
xdata.shape, ydata.shape
 
## 입력 제외 레이어의 갯수
n_h = 1
 
## x의 피처의 갯수.
n_x = 9
 
## 출력층 노드의 갯수.
n_y = 1
 
## 관측치의 갯수
m = 569 # 미사용
 
X  = tf.placeholder(tf.float64, shape = [ n_x, None ] )
y  = tf.placeholder(tf.float64, shape = [  1 , None ] )
 
W1 = tf.Variable( tf.random.normal( [ 1, n_x ], dtype = tf.float64), dtype = tf.float64 )
B1 = tf.Variable( tf.random.normal( [ 1      ], dtype = tf.float64), dtype = tf.float64 )
 
## 1번째 레이어의 입력 계산식 Z
Z1 = W1 @ X + B1
 
## 1번째 레이어의 엑티베이션 펑션
A1 = tf.sigmoid( Z1 )
 
## loss(A1, y)의 총합의 평균
# cost  = -tf.reduce_mean( y * tf.log( A1 ) + (1 - y) * tf.log( 1 - A1 ) )
 
cost  = tf.losses.sigmoid_cross_entropy(y, A1)
 
train = tf.train.GradientDescentOptimizer(0.01).minimize( cost )
 
predicted = tf.cast( A1 > 0.5, dtype = tf.float64 ) 
accuracy  = tf.reduce_mean(tf.cast(tf.equal(predicted ,y)  , dtype = tf.float64))
 
m = xdata.shape[1]
indices = np.random.permutation(m)
 
## k = 7 k-fold validation
 
k = 7
acc = []
 
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(10000):
            cv, _ = sess.run([ cost, train ], feed_dict = { X : xdata[:,train_indices], y : ydata[:,train_indices] } )
            if step % 2000 == 0:
                print(step, cv)
                    
        preds = sess.run(accuracy, feed_dict = { X : xdata[:,test_indices], y : ydata[:,test_indices] })  
        acc.append(preds)
 
 
print('acc : ', np.sum(acc) / k) ## acc :  0.9341490299823632
 
"""#2 번 문제 성능 향상"""
 
df = pd.read_csv(PATH + "house/train.csv")
features = ['sqft_living','bedrooms', 'bathrooms', 'sqft_lot','floors','zipcode','view', 'waterfront', 'grade','yr_built']
 
from sklearn.model_selection import train_test_split
train_df, test_df = train_test_split( df, test_size = 0.2, random_state = 42 )
 
import xgboost as xgb
model = xgb.XGBRegressor(max_depth = 5, n_estimators = 5000, learning_rate = 0.03)
 
model.fit(train_df[features], train_df['price'])
model.score(test_df[features], test_df['price'])  ## 0.8223037240035181