In [129]:
import pandas as pd
from sklearn.utils import shuffle
from sklearn.model_selection import train_test_split
import tensorflow as tf
from tensorflow.keras import models,layers,callbacks
from tensorflow.keras.utils import to_categorical,plot_model
import matplotlib.pyplot as plt
from sklearn import metrics
In [130]:
galaxies = pd.read_csv('./LAMOST_spectra/galaxy_standardised.csv')
qsos = pd.read_csv('./LAMOST_spectra/qso_standardised.csv')
spectras = galaxies.append(qsos)
In [131]:
spectras= shuffle(spectras)
spectras.head()
Out[131]:
0 1 2 3 4 5 6 7 8 9 ... 2541 2542 2543 2544 2545 2546 2547 2548 2549 type
183 0.520420 0.796625 1.413814 0.764956 0.518733 1.521226 1.566453 1.548385 0.599072 0.847522 ... 0.598801 0.373492 -0.824916 0.712631 0.995689 0.801263 0.814050 0.908927 0.925360 2
641 1.904963 1.733844 1.312233 1.294592 2.325209 2.242320 1.814656 1.693423 1.959508 2.122070 ... 1.326703 2.017314 0.467593 0.678183 0.910054 0.757424 0.681956 0.623207 0.697678 3
1888 2.117233 -0.117027 1.701781 1.581559 3.221272 2.307321 1.833122 2.771673 0.508816 0.976961 ... 0.749147 1.345877 -0.411363 -0.645604 -0.326204 1.337881 0.607589 0.256465 0.595038 3
756 0.642569 0.728573 0.828791 1.073927 1.248145 0.705394 0.754607 0.986046 0.810766 0.808010 ... 0.634465 0.664933 0.684013 0.667660 0.639196 0.604026 0.647809 0.671833 0.669317 2
237 -1.115247 3.226563 0.663883 0.856039 -0.075342 2.971668 3.536978 2.138347 1.860007 1.488960 ... 3.197236 3.451466 1.094434 -0.854920 -0.705939 0.053918 0.118448 0.596912 0.931616 3

5 rows × 2551 columns

In [132]:
spectras.loc[spectras['type'] == 2,'type'] = 0
spectras.loc[spectras['type'] == 3,'type'] = 1
In [133]:
spectras.head()
Out[133]:
0 1 2 3 4 5 6 7 8 9 ... 2541 2542 2543 2544 2545 2546 2547 2548 2549 type
183 0.520420 0.796625 1.413814 0.764956 0.518733 1.521226 1.566453 1.548385 0.599072 0.847522 ... 0.598801 0.373492 -0.824916 0.712631 0.995689 0.801263 0.814050 0.908927 0.925360 0
641 1.904963 1.733844 1.312233 1.294592 2.325209 2.242320 1.814656 1.693423 1.959508 2.122070 ... 1.326703 2.017314 0.467593 0.678183 0.910054 0.757424 0.681956 0.623207 0.697678 1
1888 2.117233 -0.117027 1.701781 1.581559 3.221272 2.307321 1.833122 2.771673 0.508816 0.976961 ... 0.749147 1.345877 -0.411363 -0.645604 -0.326204 1.337881 0.607589 0.256465 0.595038 1
756 0.642569 0.728573 0.828791 1.073927 1.248145 0.705394 0.754607 0.986046 0.810766 0.808010 ... 0.634465 0.664933 0.684013 0.667660 0.639196 0.604026 0.647809 0.671833 0.669317 0
237 -1.115247 3.226563 0.663883 0.856039 -0.075342 2.971668 3.536978 2.138347 1.860007 1.488960 ... 3.197236 3.451466 1.094434 -0.854920 -0.705939 0.053918 0.118448 0.596912 0.931616 1

5 rows × 2551 columns

In [5]:
# train, test = train_test_split(spectras, test_size=0.2)
# train, val = train_test_split(train, test_size=0.2)
In [134]:
X_train, X_test, y_train, y_test = train_test_split(spectras.iloc[:,0:2550],spectras['type'], test_size=0.2, random_state=0)
In [135]:
X_train, X_val, y_train, y_val = train_test_split(X_train.iloc[:,0:2550],y_train, test_size=0.2, random_state=0)
In [136]:
X_train = X_train.values
X_test = X_test.values
y_train = y_train.values
y_test = y_test.values
In [137]:
X_val = X_val.values
y_val = y_val.values
In [138]:
X_train.shape, y_train.shape
Out[138]:
((2560, 2550), (2560,))
In [148]:
y_train=to_categorical(y_train)
y_val=to_categorical(y_val)
In [149]:
y_train.shape
Out[149]:
(2560, 2)
In [150]:
X_train= X_train.reshape((-1, 2550, 1)).astype("float32")
X_test = X_test.reshape((-1, 2550, 1)).astype("float32")
In [151]:
X_val = X_val.reshape((-1, 2550, 1)).astype("float32")
In [152]:
model = models.Sequential()
model.add(layers.Conv1D(filters=32, kernel_size=5, activation='relu', input_shape=(2550,1)))
model.add(layers.MaxPooling1D(pool_size=2))
model.add(layers.Conv1D(filters=32, kernel_size=5, activation='relu'))
model.add(layers.MaxPooling1D(pool_size=2))
model.add(layers.Flatten())
model.add(layers.Dense(16, activation='relu'))
model.add(layers.Dense(2, activation='softmax'))
In [153]:
model.layers
Out[153]:
[<tensorflow.python.keras.layers.convolutional.Conv1D at 0x7f8b4830c2b0>,
 <tensorflow.python.keras.layers.pooling.MaxPooling1D at 0x7f8b4830c3c8>,
 <tensorflow.python.keras.layers.convolutional.Conv1D at 0x7f8b4830c7f0>,
 <tensorflow.python.keras.layers.pooling.MaxPooling1D at 0x7f8b48300eb8>,
 <tensorflow.python.keras.layers.core.Flatten at 0x7f8b48313748>,
 <tensorflow.python.keras.layers.core.Dense at 0x7f8b48300e80>,
 <tensorflow.python.keras.layers.core.Dense at 0x7f8b506a05c0>]
In [154]:
model.summary()

Model: "sequential_8"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv1d_16 (Conv1D)           (None, 2546, 32)          192       
_________________________________________________________________
max_pooling1d_16 (MaxPooling (None, 1273, 32)          0         
_________________________________________________________________
conv1d_17 (Conv1D)           (None, 1269, 32)          5152      
_________________________________________________________________
max_pooling1d_17 (MaxPooling (None, 634, 32)           0         
_________________________________________________________________
flatten_8 (Flatten)          (None, 20288)             0         
_________________________________________________________________
dense_16 (Dense)             (None, 16)                324624    
_________________________________________________________________
dense_17 (Dense)             (None, 2)                 34        
=================================================================
Total params: 330,002
Trainable params: 330,002
Non-trainable params: 0
_________________________________________________________________
In [155]:
model.compile(loss='binary_crossentropy', optimizer='rmsprop', metrics=['accuracy'])
In [156]:
history=model.fit(X_train,y_train,epochs=10,batch_size=32, validation_data=(X_val,y_val))

Train on 2560 samples, validate on 640 samples
Epoch 1/10
2560/2560 [==============================] - 2s 732us/sample - loss: 0.3623 - accuracy: 0.8699 - val_loss: 0.2426 - val_accuracy: 0.9250
Epoch 2/10
2560/2560 [==============================] - 1s 473us/sample - loss: 0.2184 - accuracy: 0.9395 - val_loss: 0.2520 - val_accuracy: 0.9328
Epoch 3/10
2560/2560 [==============================] - 1s 467us/sample - loss: 0.1684 - accuracy: 0.9570 - val_loss: 0.2204 - val_accuracy: 0.9344
Epoch 4/10
2560/2560 [==============================] - 1s 464us/sample - loss: 0.1373 - accuracy: 0.9621 - val_loss: 0.3544 - val_accuracy: 0.9000
Epoch 5/10
2560/2560 [==============================] - 1s 480us/sample - loss: 0.1093 - accuracy: 0.9734 - val_loss: 0.2581 - val_accuracy: 0.9359
Epoch 6/10
2560/2560 [==============================] - 1s 492us/sample - loss: 0.1002 - accuracy: 0.9758 - val_loss: 0.2983 - val_accuracy: 0.9359
Epoch 7/10
2560/2560 [==============================] - 1s 484us/sample - loss: 0.0860 - accuracy: 0.9809 - val_loss: 0.3514 - val_accuracy: 0.9250
Epoch 8/10
2560/2560 [==============================] - 1s 483us/sample - loss: 0.0690 - accuracy: 0.9812 - val_loss: 0.3000 - val_accuracy: 0.9359
Epoch 9/10
2560/2560 [==============================] - 1s 477us/sample - loss: 0.0639 - accuracy: 0.9840 - val_loss: 0.4772 - val_accuracy: 0.9141
Epoch 10/10
2560/2560 [==============================] - 1s 474us/sample - loss: 0.0483 - accuracy: 0.9895 - val_loss: 0.3928 - val_accuracy: 0.9281
In [30]:
history=model.fit(X_train,y_train,epochs=20,batch_size=32, validation_data=(X_val,y_val),callbacks=[
        callbacks.EarlyStopping(monitor="val_loss", patience=5, mode="min")
    ])

Train on 2560 samples, validate on 640 samples
Epoch 1/20
2560/2560 [==============================] - 2s 845us/sample - loss: 0.5918 - accuracy: 0.7621 - val_loss: 0.3076 - val_accuracy: 0.9219
Epoch 2/20
2560/2560 [==============================] - 1s 431us/sample - loss: 0.2895 - accuracy: 0.9023 - val_loss: 0.2532 - val_accuracy: 0.9250
Epoch 3/20
2560/2560 [==============================] - 1s 435us/sample - loss: 0.2366 - accuracy: 0.9164 - val_loss: 0.2121 - val_accuracy: 0.9312
Epoch 4/20
2560/2560 [==============================] - 1s 437us/sample - loss: 0.1995 - accuracy: 0.9312 - val_loss: 0.2778 - val_accuracy: 0.9094
Epoch 5/20
2560/2560 [==============================] - 1s 431us/sample - loss: 0.1886 - accuracy: 0.9336 - val_loss: 0.2235 - val_accuracy: 0.9344
Epoch 6/20
2560/2560 [==============================] - 1s 431us/sample - loss: 0.1466 - accuracy: 0.9523 - val_loss: 0.3196 - val_accuracy: 0.8906
Epoch 7/20
2560/2560 [==============================] - 1s 438us/sample - loss: 0.1381 - accuracy: 0.9582 - val_loss: 0.2780 - val_accuracy: 0.9172
Epoch 8/20
2560/2560 [==============================] - 1s 434us/sample - loss: 0.1231 - accuracy: 0.9582 - val_loss: 0.2912 - val_accuracy: 0.9000
In [46]:
plt.plot(history.history['accuracy'], label='accuracy')
plt.plot(history.history['val_accuracy'], label = 'val_accuracy')
plt.xlabel('Epoch')
plt.ylabel('Accuracy')
plt.ylim([0.5, 1])
plt.legend(loc='lower right')
plt.show()

test_loss, test_acc = model.evaluate(X_test,y_test, verbose=2)

800/800 - 0s - loss: 0.4506 - accuracy: 0.9262
In [49]:
print(test_acc)

0.92625
In [50]:
predictions = model.predict(X_test)
In [51]:
predictions
Out[51]:
array([[9.9657696e-01, 3.4230000e-03],
       [1.2145663e-12, 1.0000000e+00],
       [7.5070602e-05, 9.9992490e-01],
       ...,
       [4.6137668e-02, 9.5386237e-01],
       [5.4975968e-05, 9.9994504e-01],
       [1.1919570e-03, 9.9880803e-01]], dtype=float32)
In [56]:
y_pred = model.predict_classes(X_test)
In [57]:
y_pred
Out[57]:
array([0, 1, 1, 1, 1, 0, 0, 0, 1, 1, 0, 0, 0, 0, 0, 1, 0, 1, 1, 1, 0, 1,
       0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 1, 1, 0, 1, 1, 0, 1, 1, 1, 0, 1,
       0, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1, 0, 1, 1, 0, 0, 0, 0, 0, 1, 1, 0,
       1, 0, 0, 1, 1, 0, 0, 0, 1, 1, 1, 0, 0, 1, 1, 1, 1, 0, 1, 1, 0, 1,
       1, 0, 1, 0, 0, 1, 1, 1, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 1, 1, 1,
       0, 0, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 0, 0, 1, 0, 1, 0, 0, 0, 0, 1,
       1, 0, 0, 1, 0, 1, 0, 1, 1, 0, 1, 1, 1, 0, 0, 1, 1, 1, 1, 1, 1, 0,
       0, 1, 1, 1, 1, 0, 1, 0, 1, 1, 0, 0, 1, 0, 0, 1, 0, 1, 1, 1, 0, 0,
       0, 0, 1, 0, 1, 1, 0, 1, 1, 0, 1, 0, 1, 1, 1, 1, 1, 0, 0, 1, 1, 1,
       1, 0, 1, 0, 1, 0, 1, 0, 0, 1, 0, 0, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0,
       0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 0, 1, 0, 0, 1, 1, 1, 1, 0, 1,
       1, 1, 0, 0, 0, 1, 1, 0, 1, 0, 1, 1, 0, 0, 1, 1, 0, 0, 1, 1, 1, 0,
       0, 1, 1, 0, 0, 0, 1, 0, 1, 1, 1, 1, 1, 0, 0, 1, 0, 0, 1, 0, 1, 0,
       1, 1, 1, 1, 0, 0, 1, 1, 0, 0, 0, 1, 1, 1, 1, 0, 0, 0, 1, 0, 0, 1,
       1, 0, 1, 0, 0, 0, 1, 0, 0, 1, 1, 1, 1, 0, 1, 1, 1, 1, 1, 1, 0, 1,
       1, 1, 0, 0, 1, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 1, 1, 0, 1, 0,
       1, 1, 1, 0, 1, 1, 0, 0, 0, 0, 1, 1, 0, 0, 1, 0, 1, 0, 0, 1, 0, 0,
       0, 0, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 1, 1, 0, 0, 0, 1, 1,
       0, 0, 0, 1, 0, 1, 0, 1, 0, 0, 0, 1, 1, 1, 0, 0, 0, 1, 0, 1, 1, 1,
       0, 1, 1, 1, 0, 1, 0, 1, 1, 1, 0, 1, 0, 0, 1, 1, 0, 0, 1, 1, 1, 0,
       0, 0, 0, 1, 1, 1, 1, 1, 0, 1, 0, 0, 1, 1, 1, 0, 0, 0, 1, 1, 1, 1,
       0, 1, 0, 0, 1, 1, 0, 1, 0, 0, 1, 0, 1, 0, 1, 1, 1, 0, 0, 1, 0, 0,
       1, 0, 1, 1, 0, 0, 1, 0, 1, 1, 1, 0, 1, 0, 1, 0, 0, 1, 0, 0, 1, 0,
       0, 0, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 1,
       1, 0, 0, 1, 0, 1, 0, 0, 0, 0, 1, 1, 0, 1, 0, 1, 1, 1, 1, 0, 1, 1,
       1, 0, 0, 1, 1, 1, 1, 0, 1, 1, 0, 1, 1, 0, 0, 0, 1, 1, 1, 1, 1, 0,
       0, 1, 0, 0, 1, 0, 0, 1, 1, 1, 0, 1, 1, 0, 0, 1, 0, 1, 0, 0, 1, 1,
       0, 1, 1, 1, 1, 1, 0, 1, 1, 0, 1, 1, 1, 1, 0, 1, 1, 1, 0, 1, 1, 0,
       1, 0, 0, 0, 1, 1, 1, 0, 0, 1, 0, 1, 1, 0, 0, 0, 1, 0, 1, 0, 1, 0,
       0, 1, 0, 1, 1, 0, 0, 0, 0, 0, 1, 0, 1, 1, 1, 0, 1, 1, 1, 0, 0, 1,
       1, 1, 0, 0, 0, 1, 1, 1, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 1, 1,
       1, 1, 0, 1, 1, 0, 1, 0, 0, 1, 1, 1, 0, 1, 1, 1, 0, 0, 1, 0, 0, 0,
       0, 0, 0, 1, 0, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1, 0, 1, 1, 1, 1, 0, 1,
       1, 1, 1, 0, 1, 0, 1, 1, 0, 0, 1, 0, 0, 1, 1, 0, 1, 1, 1, 1, 1, 1,
       1, 0, 1, 1, 1, 1, 1, 1, 1, 0, 1, 1, 0, 1, 0, 0, 1, 0, 1, 0, 0, 1,
       1, 0, 1, 0, 0, 0, 1, 1, 0, 0, 1, 1, 1, 0, 0, 1, 0, 1, 1, 0, 0, 1,
       1, 1, 0, 1, 0, 1, 1, 1])
In [38]:
model.save('cnn_model.h5')
In [39]:
model = tf.keras.models.load_model('cnn_model.h5')
In [40]:
model.summary()

Model: "sequential_2"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv1d_4 (Conv1D)            (None, 2546, 32)          192       
_________________________________________________________________
max_pooling1d_4 (MaxPooling1 (None, 1273, 32)          0         
_________________________________________________________________
conv1d_5 (Conv1D)            (None, 1269, 32)          5152      
_________________________________________________________________
max_pooling1d_5 (MaxPooling1 (None, 634, 32)           0         
_________________________________________________________________
flatten_2 (Flatten)          (None, 20288)             0         
_________________________________________________________________
dense_4 (Dense)              (None, 16)                324624    
_________________________________________________________________
dense_5 (Dense)              (None, 2)                 34        
=================================================================
Total params: 330,002
Trainable params: 330,002
Non-trainable params: 0
_________________________________________________________________
In [60]:
metrics.roc_auc_score(y_test,predictions)
Out[60]:
0.947717731829574
In [ ]: