[케라스/텐서플로우] 주택가격 예측하기 코드

라이브러리 호출 

import tensorflow as tf
import pandas as pd
import numpy as np 
import matplotlib.pyplot as plt
import seaborn as sns

 

케라스버전 

 

데이터 불러오기 

 

from keras.models import Sequential 
from keras.layers import Dense, Activation 
from keras.datasets import boston_housing

(trainData, trainTarget),(testData, testTarget) = boston_housing.load_data()

print(trainData.shape)  # (404, 13)
print(trainTarget.shape)  # (404,)
print(testData.shape)   # (102, 13)
print(testTarget.shape)   #(102,)


# 모델 검증 데이터 분류 
ValData=trainData[304:,:]
ValTarget=trainTarget[304:]

p_trainData=trainData[:304,:]
p_trainTarget=trainTarget[:304]

 

모델 생성 

# 모델 구성 
model = Sequential()
model.add(Dense(units=30, input_dim=13,activation="relu"))
model.add(Dense(units=10,activation="relu"))
model.add(Dense(units=1))

# 모델 학습 
model.compile(loss="mse", optimizer= 'rmsprop', metrics=['mae'])

from keras.callbacks import EarlyStopping
es =EarlyStopping(patience=20)

seed= 123
np.random.seed(seed)
tf.set_random_seed(seed)

hist = model.fit(trainData,trainTarget, epochs=10, batch_size=30, callbacks=[es], 
                 validation_data=(ValData,ValTarget))

# 검증결과 출력 
print("loss:"+ str(hist.history['loss']))
print("MAE:"+ str(hist.history['mae']))

>>>
loss:[4047.806107020614, 836.5637676881091, 390.45693002832996, 249.5056139502195, 182.78083068545502, 138.7186091583554, 108.03168007878973, 88.64472017193785, 77.60753329437559, 71.6355998020361]
MAE:[56.961746, 25.148031, 16.852732, 13.033249, 10.752718, 8.986338, 7.559703, 6.517845, 6.216516, 5.886183]

 

모델 테스트 

# 모델 테스트 
res = model.evaluate(testData, testTarget, batch_size=32)

print("loss")
print(res[0])
print("mae")
print(res[1])

>>>

102/102 [==============================] - 0s 39us/step
loss
73.26150662291283
mae
6.013505935668945

모델 사용 

#7 모델 사용
xhat = testData
yhat = model.predict(testData)

%matplotlib notebook
plt.figure()
plt.plot(yhat, label = "predicted")
plt.plot(testTarget ,label = "actual")

plt.legend(prop={'size': 20})

print("Evaluate : {}".format(np.average(np.sqrt((yhat - testTarget)**2))))

 

 

텐서플로우버젼 

 

라이브러리 호출 

from keras.datasets import boston_housing

import tensorflow as tf
import pandas as pd
import numpy as np 
import matplotlib.pyplot as plt
import seaborn as sns
from tqdm import tqdm
from tensorflow.examples.tutorials.mnist import input_data

데이터 불러오기 

(trainData, trainTarget),(testData, testTarget) = boston_housing.load_data()

 

변수 및 함수 정의 

x =tf.placeholder(tf.float32,shape=[None,13])
y =tf.placeholder(tf.float32,shape=[None])

#xavier(재비어)알고리즘 기반 weight 초기화 
w1 = tf.get_variable("w1", shape=[13,30], initializer=tf.contrib.layers.xavier_initializer())
b1 = tf.Variable(tf.random_normal([30]))
l1 = tf.nn.relu(tf.matmul(x,w1)+b1)

w2 = tf.get_variable("w2", shape=[30,10], initializer=tf.contrib.layers.xavier_initializer())
b2 = tf.Variable(tf.random_normal([10]))
l2 = tf.nn.relu(tf.matmul(l1,w2)+b2)

w3 = tf.get_variable("w3", shape=[10,1], initializer=tf.contrib.layers.xavier_initializer())
b3 = tf.Variable(tf.random_normal([1]))
hf = tf.matmul(l2,w3)+b3

# 비용함수 
cost =tf.sqrt(tf.reduce_mean(tf.square(hf-y)))

# 트레이닝 함수
train = tf.train.GradientDescentOptimizer(0.01).minimize(cost)

def numIter(data):
    return int(len(data)/ batch_size )

# 정확도 
accuracy = tf.sqrt(tf.reduce_mean(tf.square(hf - y)))

 

모델 학습 및 검증 

#parameter 설정 
trainingEpochs=8
batch_size=1

sess=tf.Session()
sess.run(tf.global_variables_initializer())
for epoch in range(trainingEpochs):
    avgCost=0
    totalBatch=int(numIter(trainData)/batch_size)
    pbar=tqdm(range(totalBatch))    
    for i in pbar:  #in range(totalBatch):
#         batchX,batchY=mnist.train.next_batch(batch_size)
        cv,_,hfv=sess.run([cost,train,hf],feed_dict={x:trainData,y:trainTarget})
        avgCost+=cv/totalBatch
        pbar.set_description("cost:%f" % avgCost)
print("Evaluate:",sess.run(accuracy,feed_dict={x:testData, y:testTarget}))
>>>
Evaluate: 9.160306

 

모델 사용 

#7 모델 사용
sess= tf.Session()
sess.run(tf.global_variables_initializer())
for epoch in range(trainingEpochs):
    totalBatch = int(numIter(testData)/batch_size)
    pbar = tqdm(range(totalBatch))
    for i in pbar:
        hfv = sess.run([hf],feed_dict={x:testData})

 

예측결과 시각화 

#7 예측 시각화 
xhat = testData
yhat = hfv[0]
        
%matplotlib notebook
plt.figure()
plt.plot(yhat, label = "predicted")
plt.plot(testTarget ,label = "actual")

plt.legend(prop={'size': 20})