ex2TP1

November 5, 2024

[59]: import pandas as pd

import numpy as np
from matplotlib import pyplot as plt
%matplotlib inline
import matplotlib
matplotlib.rcParams["figure.figsize"]=(20,10)
import seaborn as sns

[60]: import sklearn

from sklearn.linear_model import LinearRegression
from sklearn.model_selection import train_test_split
from sklearn.metrics import mean_squared_error, r2_score
from sklearn.preprocessing import StandardScaler

[61]: df1=pd.read_csv("housing.csv")
df1.head()

[61]: longitude latitude housing_median_age total_rooms total_bedrooms \

0 -122.23 37.88 41.0 880.0 129.0
1 -122.22 37.86 21.0 7099.0 1106.0
2 -122.24 37.85 52.0 1467.0 190.0
3 -122.25 37.85 52.0 1274.0 235.0
4 -122.25 37.85 52.0 1627.0 280.0

population households median_income median_house_value ocean_proximity

0 322.0 126.0 8.3252 452600.0 NEAR BAY
1 2401.0 1138.0 8.3014 358500.0 NEAR BAY
2 496.0 177.0 7.2574 352100.0 NEAR BAY
3 558.0 219.0 5.6431 341300.0 NEAR BAY
4 565.0 259.0 3.8462 342200.0 NEAR BAY

[62]: df1.shape

[62]: (20640, 10)

[63]: df1.isnull().sum()

1
[63]: longitude 0
latitude 0
housing_median_age 0
total_rooms 0
total_bedrooms 207
population 0
households 0
median_income 0
median_house_value 0
ocean_proximity 0
dtype: int64

[64]: df2=df1.dropna()
df2.isnull().sum()

[64]: longitude 0
latitude 0
housing_median_age 0
total_rooms 0
total_bedrooms 0
population 0
households 0
median_income 0
median_house_value 0
ocean_proximity 0
dtype: int64

[65]: from sklearn.model_selection import train_test_split

x=df2.drop(['median_house_value'],axis=1)
y=df2['median_house_value']

[66]: x

[66]: longitude latitude housing_median_age total_rooms total_bedrooms \

0 -122.23 37.88 41.0 880.0 129.0
1 -122.22 37.86 21.0 7099.0 1106.0
2 -122.24 37.85 52.0 1467.0 190.0
3 -122.25 37.85 52.0 1274.0 235.0
4 -122.25 37.85 52.0 1627.0 280.0
… … … … … …
20635 -121.09 39.48 25.0 1665.0 374.0
20636 -121.21 39.49 18.0 697.0 150.0
20637 -121.22 39.43 17.0 2254.0 485.0
20638 -121.32 39.43 18.0 1860.0 409.0
20639 -121.24 39.37 16.0 2785.0 616.0

2
 population households median_income ocean_proximity
0 322.0 126.0 8.3252 NEAR BAY
1 2401.0 1138.0 8.3014 NEAR BAY
2 496.0 177.0 7.2574 NEAR BAY
3 558.0 219.0 5.6431 NEAR BAY
4 565.0 259.0 3.8462 NEAR BAY
… … … … …
20635 845.0 330.0 1.5603 INLAND
20636 356.0 114.0 2.5568 INLAND
20637 1007.0 433.0 1.7000 INLAND
20638 741.0 349.0 1.8672 INLAND
20639 1387.0 530.0 2.3886 INLAND

[20433 rows x 9 columns]

[67]: y

[67]: 0 452600.0
1 358500.0
2 352100.0
3 341300.0
4 342200.0
…
20635 78100.0
20636 77100.0
20637 92300.0
20638 84700.0
20639 89400.0
Name: median_house_value, Length: 20433, dtype: float64

[68]: x_train,x_test,y_train,y_test = train_test_split(x,y,test_size=0.20)

[69]: train_data=x_train.join(y_train)
#train_data
#train_data['ocean_proximity'].unique()

[70]: train_data.hist()

[70]: array([[<Axes: title={'center': 'longitude'}>,

<Axes: title={'center': 'latitude'}>,
<Axes: title={'center': 'housing_median_age'}>],
[<Axes: title={'center': 'total_rooms'}>,
<Axes: title={'center': 'total_bedrooms'}>,
<Axes: title={'center': 'population'}>],
[<Axes: title={'center': 'households'}>,
<Axes: title={'center': 'median_income'}>,
<Axes: title={'center': 'median_house_value'}>]], dtype=object)

3
[71]: train_data.corr(numeric_only=True)
plt.figure()
sns.heatmap(train_data.corr(numeric_only=True), annot=True, cmap="YlGnBu")
plt.show()

[72]: train_data['total_rooms']=np.log(train_data['total_rooms']+1)
train_data['total_bedrooms']=np.log(train_data['total_bedrooms']+1)
train_data['population']=np.log(train_data['population']+1)

4
 train_data['households']=np.log(train_data['households']+1)

[73]: train_data.hist()

[73]: array([[<Axes: title={'center': 'longitude'}>,

<Axes: title={'center': 'latitude'}>,
<Axes: title={'center': 'housing_median_age'}>],
[<Axes: title={'center': 'total_rooms'}>,
<Axes: title={'center': 'total_bedrooms'}>,
<Axes: title={'center': 'population'}>],
[<Axes: title={'center': 'households'}>,
<Axes: title={'center': 'median_income'}>,
<Axes: title={'center': 'median_house_value'}>]], dtype=object)

[74]: train_data.ocean_proximity.value_counts()

[74]: ocean_proximity
<1H OCEAN 7217
INLAND 5170
NEAR OCEAN 2118
NEAR BAY 1837
ISLAND 4
Name: count, dtype: int64

[75]: #pd.get_dummies(train_data.ocean_proximity).astype(int)
#train_data.join(pd.get_dummies(train_data.ocean_proximity.astype(int))).
↪drop(['ocean_proximity'],axis=1)

dummies = pd.get_dummies(train_data['ocean_proximity']).astype(int)

5
 train_data = train_data.join(dummies).drop(['ocean_proximity'], axis=1)
train_data

[75]: longitude latitude housing_median_age total_rooms total_bedrooms \

8424 -118.36 33.93 27.0 8.399760 7.116394
19741 -122.57 39.90 15.0 8.262043 6.698268
19155 -122.71 38.37 16.0 7.764721 5.846439
7751 -118.15 33.92 28.0 6.946014 5.533389
17400 -120.44 34.93 15.0 6.767343 5.501258
… … … … … …
17208 -119.72 34.43 36.0 7.053586 5.736572
19855 -119.45 36.35 22.0 7.509335 5.811141
19380 -120.85 37.77 52.0 6.079933 4.406719
17254 -119.71 34.42 39.0 7.067320 5.777652
8760 -118.44 33.81 33.0 8.292799 6.898715

population households median_income median_house_value <1H OCEAN \

8424 8.114025 7.015712 3.1656 204500.0 1
19741 7.437206 6.442540 2.4555 55600.0 0
19155 6.922644 5.855072 5.6018 253000.0 1
7751 6.816736 5.505332 2.5875 161200.0 1
17400 7.033506 5.537334 2.0995 87500.0 1
… … … … … …
17208 6.257668 5.720312 2.6014 320600.0 1
19855 6.981935 5.645447 2.3365 69600.0 0
19380 5.288267 4.234107 1.8625 85400.0 0
17254 6.408529 5.758902 2.1600 259100.0 1
8760 7.407318 6.837333 5.0106 500001.0 0

INLAND ISLAND NEAR BAY NEAR OCEAN

8424 0 0 0 0
19741 1 0 0 0
19155 0 0 0 0
7751 0 0 0 0
17400 0 0 0 0
… … … … …
17208 0 0 0 0
19855 1 0 0 0
19380 1 0 0 0
17254 0 0 0 0
8760 0 0 0 1

[16346 rows x 14 columns]

[76]: plt.figure()
sns.heatmap(train_data.corr(),annot=True,cmap="YlGnBu")
plt.show()

6
[77]: plt.figure()
sns.scatterplot(x='longitude',y='latitude',data=train_data,␣
↪hue="median_house_value",palette="coolwarm")

[77]: <Axes: xlabel='longitude', ylabel='latitude'>

7
[78]: train_data['bedrooms_ratio']=train_data['total_bedrooms'] /␣
↪train_data['total_rooms']

train_data['household_rooms']=train_data['total_rooms'] /␣
↪train_data['households']

[79]: plt.figure()
sns.heatmap(train_data.corr(),annot=True,cmap="YlGnBu")
plt.show()

[80]: x_train=train_data.drop(['median_house_value'],axis=1)
y_train=train_data['median_house_value']

#normalisation des donnees

scaler = StandardScaler()
x_train_scaled = scaler.fit_transform(x_train)

model=LinearRegression()
model.fit(x_train_scaled,y_train_scaled)

[80]: LinearRegression()

[81]: test_data=x_test.join(y_test)
test_data['total_rooms']=np.log(test_data['total_rooms']+1)
test_data['total_bedrooms']=np.log(test_data['total_bedrooms']+1)

8
 test_data['population']=np.log(test_data['population']+1)
test_data['households']=np.log(test_data['households']+1)

test_data = test_data.join(pd.get_dummies(test_data['ocean_proximity']).
↪astype(int)).drop(['ocean_proximity'], axis=1)

test_data['bedrooms_ratio']=test_data['total_bedrooms'] /␣
↪test_data['total_rooms']

test_data['household_rooms']=test_data['total_rooms'] / test_data['households']

#pd.get_dummies(test_data['ocean_proximity'])

[82]: test_data

[82]: longitude latitude housing_median_age total_rooms total_bedrooms \

10713 -117.84 33.66 5.0 6.501290 5.147494
20468 -118.71 34.27 26.0 6.898715 5.411646
17148 -122.20 37.43 38.0 8.196161 6.270988
7645 -118.27 33.81 10.0 7.540090 6.349139
17440 -120.44 34.66 22.0 8.080856 6.309918
… … … … … …
14905 -117.06 32.60 24.0 6.993015 5.594711
15752 -122.44 37.77 52.0 8.153637 6.694562
8273 -118.16 33.77 29.0 8.032360 6.668228
1746 -122.35 37.96 34.0 7.264730 5.817111
15696 -122.45 37.79 52.0 7.458763 6.180017

population households median_income median_house_value <1H OCEAN \

10713 5.953243 5.147494 4.5833 230400.0 1
20468 6.579251 5.451038 3.1630 179400.0 1
17148 7.208600 6.278521 7.3681 500001.0 0
7645 7.478735 6.317165 3.9286 114000.0 1
17440 7.461640 6.366470 4.5417 142400.0 0
… … … … … …
14905 6.999422 5.509388 2.4191 107300.0 0
15752 7.325808 6.656727 3.6186 500001.0 0
8273 7.286876 6.602588 2.8750 232500.0 0
1746 7.149132 5.768321 2.5461 93900.0 0
15696 6.595781 6.063785 1.4804 425000.0 0

INLAND ISLAND NEAR BAY NEAR OCEAN bedrooms_ratio household_rooms

10713 0 0 0 0 0.791765 1.263001
20468 0 0 0 0 0.784443 1.265578
17148 0 0 0 1 0.765113 1.305429
7645 0 0 0 0 0.842051 1.193588
17440 0 0 0 1 0.780848 1.269284
… … … … … … …
14905 0 0 0 1 0.800043 1.269291

9
 15752 0 0 1 0 0.821052 1.224872
8273 0 0 0 1 0.830170 1.216547
1746 0 0 1 0 0.800733 1.259419
15696 0 0 1 0 0.828558 1.230051

[4087 rows x 16 columns]

[136]: x_test,y_test=test_data.
↪drop(['median_house_value'],axis=1),test_data['median_house_value']

x_test_scaled = scaler.transform(x_test)
#y_test_scaled = scaler.transform(y_test.values.reshape(-1, 1))
'''
y_scaler = StandardScaler()
y_train_scaled = y_scaler.fit_transform(y_train.values.reshape(-1, 1))
y_test_scaled = y_scaler.transform(y_test.values.reshape(-1, 1))
'''

[84]: x_test

[84]: longitude latitude housing_median_age total_rooms total_bedrooms \

10713 -117.84 33.66 5.0 6.501290 5.147494
20468 -118.71 34.27 26.0 6.898715 5.411646
17148 -122.20 37.43 38.0 8.196161 6.270988
7645 -118.27 33.81 10.0 7.540090 6.349139
17440 -120.44 34.66 22.0 8.080856 6.309918
… … … … … …
14905 -117.06 32.60 24.0 6.993015 5.594711
15752 -122.44 37.77 52.0 8.153637 6.694562
8273 -118.16 33.77 29.0 8.032360 6.668228
1746 -122.35 37.96 34.0 7.264730 5.817111
15696 -122.45 37.79 52.0 7.458763 6.180017

population households median_income <1H OCEAN INLAND ISLAND \

10713 5.953243 5.147494 4.5833 1 0 0
20468 6.579251 5.451038 3.1630 1 0 0
17148 7.208600 6.278521 7.3681 0 0 0
7645 7.478735 6.317165 3.9286 1 0 0
17440 7.461640 6.366470 4.5417 0 0 0
… … … … … … …
14905 6.999422 5.509388 2.4191 0 0 0
15752 7.325808 6.656727 3.6186 0 0 0
8273 7.286876 6.602588 2.8750 0 0 0
1746 7.149132 5.768321 2.5461 0 0 0
15696 6.595781 6.063785 1.4804 0 0 0

NEAR BAY NEAR OCEAN bedrooms_ratio household_rooms

10713 0 0 0.791765 1.263001

10
 20468 0 0 0.784443 1.265578
17148 0 1 0.765113 1.305429
7645 0 0 0.842051 1.193588
17440 0 1 0.780848 1.269284
… … … … …
14905 0 1 0.800043 1.269291
15752 1 0 0.821052 1.224872
8273 0 1 0.830170 1.216547
1746 1 0 0.800733 1.259419
15696 1 0 0.828558 1.230051

[4087 rows x 15 columns]

[138]: model.score(x_train_scaled, y_train_scaled)

---------------------------------------------------------------------------
AttributeError Traceback (most recent call last)
Cell In[138], line 1
----> 1 model.score(x_train_scaled, y_train_scaled)

AttributeError: 'Sequential' object has no attribute 'score'

[86]: y_pred = model.predict(x_test_scaled)

[87]: mse = mean_squared_error(y_test, y_pred)

mse

[87]: 57926172223.61929

[88]: from sklearn.ensemble import RandomForestRegressor

forest=RandomForestRegressor()
forest.fit(x_train,y_train)

[88]: RandomForestRegressor()

[89]: forest.score(x_test,y_test)

[89]: 0.8146615773444194

[91]: import keras

from keras.models import Sequential
from keras.layers import Dense

[92]: #definir le modele

model = Sequential([Dense(1, input_dim=x_train.shape[1], activation='linear')])
model.compile(optimizer='adam', loss='mean_squared_error')
model.summary()

11
 C:\Users\king\anaconda3\Lib\site-packages\keras\src\layers\core\dense.py:87:
UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When
using Sequential models, prefer using an `Input(shape)` object as the first
layer in the model instead.
super().__init__(activity_regularizer=activity_regularizer, **kwargs)
Model: "sequential"

����������������������������������������������������������������������������
� Layer (type) � Output Shape � Param # �
����������������������������������������������������������������������������
� dense (Dense) � (None, 1) � 16 �
����������������������������������������������������������������������������

Total params: 16 (64.00 B)

Trainable params: 16 (64.00 B)

Non-trainable params: 0 (0.00 B)

[93]: train_5 = model.fit(x_train, y_train, validation_data=(x_test, y_test),␣

↪epochs=5)

# Entraînement pour 10 epochs

train_10 = model.fit(x_train, y_train, validation_data=(x_test, y_test),␣
↪epochs=10)

# Entraînement pour 15 epochs

train_15 = model.fit(x_train, y_train, validation_data=(x_test, y_test),␣
↪epochs=15)

Epoch 1/5
511/511 �������������������� 3s 4ms/step -
loss: 55905394688.0000 - val_loss: 57862823936.0000
Epoch 2/5
511/511 �������������������� 3s 6ms/step -
loss: 55015342080.0000 - val_loss: 57816014848.0000
Epoch 3/5
511/511 �������������������� 2s 4ms/step -
loss: 55397969920.0000 - val_loss: 57769230336.0000
Epoch 4/5
511/511 �������������������� 3s 5ms/step -
loss: 55470444544.0000 - val_loss: 57722441728.0000
Epoch 5/5
511/511 �������������������� 2s 3ms/step -

12
loss: 55485304832.0000 - val_loss: 57675702272.0000
Epoch 1/10
511/511 �������������������� 2s 3ms/step -
loss: 56076288000.0000 - val_loss: 57629020160.0000
Epoch 2/10
511/511 �������������������� 3s 6ms/step -
loss: 55215300608.0000 - val_loss: 57582419968.0000
Epoch 3/10
511/511 �������������������� 1s 2ms/step -
loss: 55182422016.0000 - val_loss: 57535737856.0000
Epoch 4/10
511/511 �������������������� 1s 3ms/step -
loss: 54860214272.0000 - val_loss: 57489137664.0000
Epoch 5/10
511/511 �������������������� 1s 3ms/step -
loss: 55305433088.0000 - val_loss: 57442570240.0000
Epoch 6/10
511/511 �������������������� 2s 3ms/step -
loss: 54938718208.0000 - val_loss: 57396002816.0000
Epoch 7/10
511/511 �������������������� 2s 3ms/step -
loss: 55876726784.0000 - val_loss: 57349505024.0000
Epoch 8/10
511/511 �������������������� 2s 3ms/step -
loss: 55152832512.0000 - val_loss: 57302986752.0000
Epoch 9/10
511/511 �������������������� 1s 3ms/step -
loss: 54960181248.0000 - val_loss: 57256488960.0000
Epoch 10/10
511/511 �������������������� 3s 6ms/step -
loss: 56382586880.0000 - val_loss: 57210064896.0000
Epoch 1/15
511/511 �������������������� 2s 3ms/step -
loss: 55094513664.0000 - val_loss: 57163624448.0000
Epoch 2/15
511/511 �������������������� 2s 3ms/step -
loss: 55323426816.0000 - val_loss: 57117188096.0000
Epoch 3/15
511/511 �������������������� 3s 5ms/step -
loss: 55270674432.0000 - val_loss: 57070817280.0000
Epoch 4/15
511/511 �������������������� 1s 3ms/step -
loss: 53296803840.0000 - val_loss: 57024409600.0000
Epoch 5/15
511/511 �������������������� 2s 3ms/step -
loss: 54269583360.0000 - val_loss: 56978063360.0000
Epoch 6/15
511/511 �������������������� 2s 3ms/step -

13
 loss: 54459072512.0000 - val_loss: 56931782656.0000
Epoch 7/15
511/511 �������������������� 2s 3ms/step -
loss: 54719565824.0000 - val_loss: 56885473280.0000
Epoch 8/15
511/511 �������������������� 2s 4ms/step -
loss: 53834076160.0000 - val_loss: 56839192576.0000
Epoch 9/15
511/511 �������������������� 2s 4ms/step -
loss: 54277263360.0000 - val_loss: 56793022464.0000
Epoch 10/15
511/511 �������������������� 3s 5ms/step -
loss: 54326079488.0000 - val_loss: 56746799104.0000
Epoch 11/15
511/511 �������������������� 2s 4ms/step -
loss: 55075880960.0000 - val_loss: 56700678144.0000
Epoch 12/15
511/511 �������������������� 2s 3ms/step -
loss: 54168629248.0000 - val_loss: 56654499840.0000
Epoch 13/15
511/511 �������������������� 2s 4ms/step -
loss: 55217995776.0000 - val_loss: 56608378880.0000
Epoch 14/15
511/511 �������������������� 2s 3ms/step -
loss: 54757396480.0000 - val_loss: 56562302976.0000
Epoch 15/15
511/511 �������������������� 2s 3ms/step -
loss: 54337257472.0000 - val_loss: 56516194304.0000

[94]: def plot_performance(train, epochs):

plt.plot(train.history['loss'], label='train loss')
plt.plot(train.history['val_loss'], label='validation loss')
plt.title(f'Loss over {epochs} epochs')
plt.xlabel('Epochs')
plt.ylabel('Mean Squared Error')
plt.legend()
plt.show()

[95]: plot_performance(train_5, 5)
plot_performance(train_10, 10)
plot_performance(train_15, 15)

14
15
[96]: model.summary()

Model: "sequential"

����������������������������������������������������������������������������
� Layer (type) � Output Shape � Param # �
����������������������������������������������������������������������������
� dense (Dense) � (None, 1) � 16 �
����������������������������������������������������������������������������

Total params: 50 (204.00 B)

Trainable params: 16 (64.00 B)

Non-trainable params: 0 (0.00 B)

Optimizer params: 34 (140.00 B)

[ ]:

16