In [1]:

    

import keras
import numpy as np
import pandas as pd

from keras.layers import Input, Dense
from keras.models import Model, Sequential

from sklearn.preprocessing import StandardScaler, MinMaxScaler, Normalizer


    

    




Using TensorFlow backend.

In [2]:

    

df = pd.read_csv("../housing-conditions-in-copenhagen/data.csv",sep="\s+")


    

In [3]:

    

df.head()


    

    
Out[3]:







  

    

      

      
housing
      
influence
      
contact_with_neighbours
      
satisfaction
      
n
    
  
  

    

      
1
      
1
      
1
      
1
      
1
      
21
    
    

      
2
      
1
      
1
      
1
      
2
      
21
    
    

      
3
      
1
      
1
      
1
      
3
      
28
    
    

      
4
      
1
      
1
      
2
      
1
      
14
    
    

      
5
      
1
      
1
      
2
      
2
      
19
    
  

In [4]:

    

df.describe()


    

    
Out[4]:







  

    

      

      
housing
      
influence
      
contact_with_neighbours
      
satisfaction
      
n
    
  
  

    

      
count
      
72.00000
      
72.000000
      
72.000000
      
72.000000
      
72.000000
    
    

      
mean
      
2.50000
      
2.000000
      
1.500000
      
2.000000
      
23.347222
    
    

      
std
      
1.12588
      
0.822226
      
0.503509
      
0.822226
      
17.666041
    
    

      
min
      
1.00000
      
1.000000
      
1.000000
      
1.000000
      
3.000000
    
    

      
25%
      
1.75000
      
1.000000
      
1.000000
      
1.000000
      
10.000000
    
    

      
50%
      
2.50000
      
2.000000
      
1.500000
      
2.000000
      
19.500000
    
    

      
75%
      
3.25000
      
3.000000
      
2.000000
      
3.000000
      
31.750000
    
    

      
max
      
4.00000
      
3.000000
      
2.000000
      
3.000000
      
86.000000
    
  

In [5]:

    

df_new = pd.concat([df,pd.get_dummies(df["housing"],prefix="housing")],axis=1).drop(["housing"],axis=1)
df_new = pd.concat([df_new,pd.get_dummies(df["influence"],prefix="influence")],axis=1).drop(["influence"],axis=1)
df_new = pd.concat([df_new,pd.get_dummies(df["satisfaction"],prefix="satisfaction")],axis=1).drop(["satisfaction"],axis=1)


    

In [6]:

    

df_new["contact_with_neighbours"] = df_new["contact_with_neighbours"].map(lambda v: 0 if v == 1 else 1)

df_new = df_new.drop(["n"])


    

In [7]:

    

df_new.head()


    

    
Out[7]:







  

    

      

      
contact_with_neighbours
      
n
      
housing_1
      
housing_2
      
housing_3
      
housing_4
      
influence_1
      
influence_2
      
influence_3
      
satisfaction_1
      
satisfaction_2
      
satisfaction_3
    
  
  

    

      
1
      
0
      
21
      
1
      
0
      
0
      
0
      
1
      
0
      
0
      
1
      
0
      
0
    
    

      
2
      
0
      
21
      
1
      
0
      
0
      
0
      
1
      
0
      
0
      
0
      
1
      
0
    
    

      
3
      
0
      
28
      
1
      
0
      
0
      
0
      
1
      
0
      
0
      
0
      
0
      
1
    
    

      
4
      
1
      
14
      
1
      
0
      
0
      
0
      
1
      
0
      
0
      
1
      
0
      
0
    
    

      
5
      
1
      
19
      
1
      
0
      
0
      
0
      
1
      
0
      
0
      
0
      
1
      
0
    
  

In [89]:

    

y = df_new["contact_with_neighbours"].values.reshape(-1,1)
X = df_new.drop(["contact_with_neighbours"], axis=1).values


    

In [90]:

    

# sc = StandardScaler()
# X = sc.fit_transform(X)


    

In [91]:

    

y.shape, X.shape


    

    
Out[91]:





((72, 1), (72, 11))

In [92]:

    

indices = np.arange(72)
np.random.shuffle(indices)
num_validation_samples = int(0.2*72)

y = y[indices]
X = X[indices]

X_train = X[:-num_validation_samples]
y_train = y[:-num_validation_samples]
X_val = X[-num_validation_samples:]
y_val = y[-num_validation_samples:]


    

In [93]:

    

X_train.shape,y_train.shape,X_val.shape, y_val.shape


    

    
Out[93]:





((58, 11), (58, 1), (14, 11), (14, 1))

In [99]:

    

from keras.regularizers import l1_l2
reg = l1_l2(l1=0.01, l2=0.01)

model = Sequential()
model.add(Dense(1,
                init='glorot_uniform',
                activation='sigmoid', 
                input_dim=X.shape[1]))


rmsprop = keras.optimizers.RMSprop(lr=0.01)# y = sc.fit_transform(y)
sgd = keras.optimizers.SGD(lr=0.1)
model.compile(optimizer=rmsprop, loss='binary_crossentropy')

early_stopping = keras.callbacks.EarlyStopping(monitor='val_loss', min_delta=0, patience=5, verbose=0, mode='auto')

model.fit(X_train, 
          y_train,
          epochs=30, 
          validation_data=(X_val, y_val),
          callbacks=[early_stopping])


    

    




/home/felipe/tf-venv3/lib/python3.5/site-packages/ipykernel_launcher.py:8: UserWarning: Update your `Dense` call to the Keras 2 API: `Dense(1, kernel_initializer="glorot_uniform", activation="sigmoid", input_dim=11)`
  






    




Train on 58 samples, validate on 14 samples
Epoch 1/30
58/58 [==============================] - 0s - loss: 3.6872 - val_loss: 4.8987
Epoch 2/30
58/58 [==============================] - 0s - loss: 3.1931 - val_loss: 4.3503
Epoch 3/30
58/58 [==============================] - 0s - loss: 2.7570 - val_loss: 3.8347
Epoch 4/30
58/58 [==============================] - 0s - loss: 2.4010 - val_loss: 3.3106
Epoch 5/30
58/58 [==============================] - 0s - loss: 2.0814 - val_loss: 2.8140
Epoch 6/30
58/58 [==============================] - 0s - loss: 1.7635 - val_loss: 2.3486
Epoch 7/30
58/58 [==============================] - 0s - loss: 1.4866 - val_loss: 1.9665
Epoch 8/30
58/58 [==============================] - 0s - loss: 1.2881 - val_loss: 1.6689
Epoch 9/30
58/58 [==============================] - 0s - loss: 1.1107 - val_loss: 1.3813
Epoch 10/30
58/58 [==============================] - 0s - loss: 0.9732 - val_loss: 1.1473
Epoch 11/30
58/58 [==============================] - 0s - loss: 0.8849 - val_loss: 0.9791
Epoch 12/30
58/58 [==============================] - 0s - loss: 0.8177 - val_loss: 0.8943
Epoch 13/30
58/58 [==============================] - 0s - loss: 0.8037 - val_loss: 0.8291
Epoch 14/30
58/58 [==============================] - 0s - loss: 0.7842 - val_loss: 0.8239
Epoch 15/30
58/58 [==============================] - 0s - loss: 0.7753 - val_loss: 0.8058
Epoch 16/30
58/58 [==============================] - 0s - loss: 0.7701 - val_loss: 0.8029
Epoch 17/30
58/58 [==============================] - 0s - loss: 0.7877 - val_loss: 0.7680
Epoch 18/30
58/58 [==============================] - 0s - loss: 0.7633 - val_loss: 0.7721
Epoch 19/30
58/58 [==============================] - 0s - loss: 0.7587 - val_loss: 0.7935
Epoch 20/30
58/58 [==============================] - 0s - loss: 0.7576 - val_loss: 0.8060
Epoch 21/30
58/58 [==============================] - 0s - loss: 0.7571 - val_loss: 0.8120
Epoch 22/30
58/58 [==============================] - 0s - loss: 0.7453 - val_loss: 0.7862
Epoch 23/30
58/58 [==============================] - 0s - loss: 0.7415 - val_loss: 0.7701






    
Out[99]:





<keras.callbacks.History at 0x7f461326bdd8>

In [ ]: