trainer.py

import keras
from keras import layers
from keras.layers.core import Dense, Activation
from keras.models import Sequential
from keras.callbacks import EarlyStopping
from keras.callbacks import ModelCheckpoint
import pandas as pd
from sklearn.model_selection import train_test_split
import numpy as np
from sklearn.metrics import accuracy_score
from keras.layers import LSTM
from sklearn import preprocessing

# Convert a Pandas dataframe to the x,y inputs that TensorFlow needs
def to_xy(df, target):
    result = []
    for x in df.columns:
        if x != target:
            result.append(x)
    # find out the type of the target column.  Is it really this hard? :(
    target_type = df[target].dtypes
    target_type = target_type[0] if hasattr(target_type, '__iter__') else target_type
    # Encode to int for classification, float otherwise. TensorFlow likes 32 bits.
    if target_type in (np.int64, np.int32):
        # Classification
        dummies = pd.get_dummies(df[target])
        return df.as_matrix(result).astype(np.float32), dummies.as_matrix().astype(np.float32)
    else:
        # Regression
        return df.as_matrix(result).astype(np.float32), df.as_matrix([target]).astype(np.float32)
    
# Encode text values to indexes(i.e. [1],[2],[3] for red,green,blue).
def encode_text_index(df, name):
    le = preprocessing.LabelEncoder()
    df[name] = le.fit_transform(df[name])
    return le.classes_
  
X,y = to_xy(df,"Actual")

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.20)

model = Sequential()
model.add(Dense(80, activation='relu', input_dim=X_train.shape[1]))
model.add(Dense(80, activation='relu'))
model.add(Dense(80, activation='relu'))
model.add(Dense(80, activation='relu'))
model.add(Dense(80, activation='relu'))
model.add(Dense(y_train.shape[1],activation='softmax'))

#model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])
#model.fit(X_train, y_train, epochs=1000)

model.compile(loss='categorical_crossentropy', optimizer='adam')
monitor = EarlyStopping(monitor='val_loss', min_delta=1e-2, patience=25, verbose=1, mode='auto')
checkpointer = ModelCheckpoint(filepath="best_weights.hdf5", verbose=0, save_best_only=True) # save best model

model.fit(X_train, y_train, validation_data=(X_test, y_test), callbacks=[monitor,checkpointer], verbose=2, epochs=1000)
model.load_weights('best_weights.hdf5') # load weights from best model

# Train on 2362 samples, validate on 591 samples
# Epoch 1/1000
#  - 3s - loss: 4.0986 - val_loss: 3.2790
# Epoch 2/1000
#  - 0s - loss: 1.9809 - val_loss: 0.4511
# Epoch 3/1000
#  - 0s - loss: 0.8889 - val_loss: 0.3435
# Epoch 4/1000
#  - 0s - loss: 0.8331 - val_loss: 0.4944
# Epoch 5/1000
#  - 0s - loss: 0.7122 - val_loss: 2.7115
# Epoch 6/1000
#  - 0s - loss: 2.3070 - val_loss: 2.1721
# Epoch 7/1000
#  - 0s - loss: 0.7217 - val_loss: 0.3364
# Epoch 8/1000
#  - 0s - loss: 0.4693 - val_loss: 0.2795
# Epoch 9/1000
#  - 0s - loss: 1.1468 - val_loss: 0.2923
# Epoch 10/1000
#  - 0s - loss: 0.3399 - val_loss: 0.2821
# Epoch 11/1000
#  - 0s - loss: 0.6785 - val_loss: 0.6393
# Epoch 12/1000
#  - 0s - loss: 0.4028 - val_loss: 0.3180
# Epoch 13/1000
#  - 0s - loss: 0.3777 - val_loss: 0.9051
# Epoch 14/1000
#  - 0s - loss: 0.3900 - val_loss: 0.2362
# Epoch 15/1000
#  - 0s - loss: 0.3052 - val_loss: 0.3325
# Epoch 16/1000
#  - 0s - loss: 0.1849 - val_loss: 0.5999
# Epoch 17/1000
#  - 0s - loss: 0.3221 - val_loss: 0.1700
# Epoch 18/1000
#  - 0s - loss: 0.3261 - val_loss: 0.5254
# Epoch 19/1000
#  - 0s - loss: 0.3268 - val_loss: 0.2400
# Epoch 20/1000
#  - 0s - loss: 0.1985 - val_loss: 0.1696
# Epoch 21/1000
#  - 0s - loss: 0.3524 - val_loss: 0.1792
# Epoch 22/1000
#  - 0s - loss: 0.3941 - val_loss: 0.2260
# Epoch 23/1000
#  - 0s - loss: 0.1884 - val_loss: 0.1476
# Epoch 24/1000
#  - 0s - loss: 0.1158 - val_loss: 0.1179
# Epoch 25/1000
#  - 0s - loss: 0.2895 - val_loss: 0.1734
# Epoch 26/1000
#  - 0s - loss: 0.2313 - val_loss: 0.1787
# Epoch 27/1000
#  - 0s - loss: 0.3110 - val_loss: 0.1681
# Epoch 28/1000
#  - 1s - loss: 0.1183 - val_loss: 0.1103
# Epoch 29/1000
#  - 1s - loss: 0.1586 - val_loss: 0.1469
# Epoch 30/1000
#  - 0s - loss: 0.0938 - val_loss: 0.1351
# Epoch 31/1000
#  - 0s - loss: 0.1677 - val_loss: 0.1374
# Epoch 32/1000
#  - 1s - loss: 0.1437 - val_loss: 0.1216
# Epoch 33/1000
#  - 1s - loss: 0.8587 - val_loss: 0.1599
# Epoch 34/1000
#  - 1s - loss: 0.1596 - val_loss: 0.1297
# Epoch 35/1000
#  - 1s - loss: 0.3959 - val_loss: 0.1348
# Epoch 36/1000
#  - 0s - loss: 0.2298 - val_loss: 0.2588
# Epoch 37/1000
#  - 0s - loss: 0.1740 - val_loss: 0.2553
# Epoch 38/1000
#  - 0s - loss: 0.1085 - val_loss: 0.1412
# Epoch 39/1000
#  - 1s - loss: 0.0834 - val_loss: 0.0953
# Epoch 40/1000
#  - 0s - loss: 0.1084 - val_loss: 0.1288
# Epoch 41/1000
#  - 0s - loss: 0.0784 - val_loss: 0.1193
# Epoch 42/1000
#  - 0s - loss: 0.0904 - val_loss: 0.1151
# Epoch 43/1000
#  - 0s - loss: 0.2667 - val_loss: 0.3757
# Epoch 44/1000
#  - 0s - loss: 0.1796 - val_loss: 0.1106
# Epoch 45/1000
#  - 1s - loss: 0.1933 - val_loss: 0.3802
# Epoch 46/1000
#  - 0s - loss: 0.2617 - val_loss: 0.3527
# Epoch 47/1000
#  - 1s - loss: 0.1542 - val_loss: 0.4486
# Epoch 48/1000
#  - 0s - loss: 0.1990 - val_loss: 0.1225
# Epoch 49/1000
#  - 1s - loss: 0.7622 - val_loss: 0.2424
# Epoch 50/1000
#  - 0s - loss: 0.2389 - val_loss: 0.1226
# Epoch 51/1000
#  - 0s - loss: 0.0747 - val_loss: 0.0958
# Epoch 52/1000
#  - 1s - loss: 0.1290 - val_loss: 0.2079
# Epoch 53/1000
#  - 0s - loss: 0.0889 - val_loss: 0.1068
# Epoch 54/1000
#  - 1s - loss: 0.0584 - val_loss: 0.0944
# Epoch 55/1000
#  - 1s - loss: 0.2151 - val_loss: 0.1194
# Epoch 56/1000
#  - 1s - loss: 0.1633 - val_loss: 0.0893
# Epoch 57/1000
#  - 1s - loss: 0.0686 - val_loss: 0.0960
# Epoch 58/1000
#  - 0s - loss: 0.0607 - val_loss: 0.0884
# Epoch 59/1000
#  - 1s - loss: 0.0651 - val_loss: 0.1915
# Epoch 60/1000
#  - 0s - loss: 0.0562 - val_loss: 0.0995
# Epoch 61/1000
#  - 1s - loss: 1.9859 - val_loss: 5.9858
# Epoch 62/1000
#  - 0s - loss: 2.9463 - val_loss: 0.9567
# Epoch 63/1000
#  - 0s - loss: 1.5587 - val_loss: 0.1932
# Epoch 64/1000
#  - 0s - loss: 1.1365 - val_loss: 2.4021
# Epoch 00064: early stopping

pred = model.predict(X_test)

print("Shape: {}".format(pred.shape))
print(pred)

# Shape: (591, 2)
# [[1.1338343e-02 9.8866165e-01]
#  [0.0000000e+00 1.0000000e+00]
#  [1.4060745e-05 9.9998593e-01]
#  ...
#  [7.8182465e-01 2.1817538e-01]
#  [0.0000000e+00 1.0000000e+00]
#  [3.2087931e-04 9.9967909e-01]]

predict_classes = np.argmax(pred,axis=1)
print("Predictions: {}".format(predict_classes))
print("Expected: {}".format(np.argmax(y_test, axis=1)))
y_test_arg = np.argmax(y_test, axis=1)

# Predictions: [1 1 1 1 0 1 0 1 0 0 0 0 1 0 1 1 1 1 0 1 0 0 0 1 1 0 1 0 0 1 0 0 0 1 0 1 1
#  1 0 1 0 1 1 0 0 0 1 0 0 1 0 1 0 0 1 1 0 0 0 1 0 0 1 1 1 0 0 1 1 1 1 0 1 0
#  0 0 1 1 1 0 1 0 1 1 0 0 0 0 1 0 1 0 1 0 1 1 1 1 0 1 1 0 1 1 0 0 1 1 1 1 1
#  1 1 1 1 0 0 1 1 1 0 1 1 1 0 1 0 0 0 0 1 1 0 1 1 1 1 1 0 1 1 0 1 0 0 0 0 1
#  1 1 1 1 0 1 1 1 0 0 0 1 0 1 0 1 0 1 1 1 1 1 1 1 1 0 0 1 0 1 0 0 1 0 0 0 1
#  1 1 0 1 1 1 1 0 0 0 1 1 0 1 1 1 0 0 1 1 0 1 1 1 0 1 1 1 1 1 0 1 0 1 1 0 0
#  0 0 1 1 1 0 1 1 0 1 1 1 1 1 1 0 0 1 1 1 1 0 1 1 1 0 1 1 1 0 1 1 0 1 1 1 0
#  0 1 1 0 1 0 1 0 0 0 0 1 0 1 0 1 1 1 1 0 1 0 1 0 1 1 0 1 1 1 1 1 0 1 0 1 0
#  1 1 1 1 0 1 0 1 1 1 1 1 0 1 1 0 1 0 0 1 1 1 1 1 0 0 0 1 1 0 1 0 1 1 1 1 1
#  1 1 1 0 1 0 1 0 1 1 1 1 0 1 0 0 0 1 1 1 1 0 0 1 1 1 0 1 0 0 1 1 1 0 1 0 0
#  1 1 1 1 0 0 1 1 1 1 1 1 0 1 1 1 1 0 0 0 1 0 1 0 0 1 0 1 1 0 1 1 1 1 0 0 1
#  1 0 1 1 1 1 1 0 0 1 1 0 1 1 0 1 0 0 1 1 0 1 1 1 1 1 0 1 0 1 1 1 1 1 1 0 0
#  0 1 0 1 0 1 0 0 1 1 0 1 0 1 1 0 1 1 1 1 1 1 0 0 0 0 1 1 1 1 1 0 1 1 1 1 1
#  0 1 1 0 1 0 0 1 1 1 1 0 0 0 0 1 1 0 1 0 0 0 0 0 0 1 1 1 1 0 1 1 1 1 0 1 1
#  1 1 1 0 1 1 0 1 0 1 0 1 1 0 1 0 1 1 1 0 0 1 0 0 1 1 1 1 1 1 1 1 1 1 1 0 0
#  0 0 0 1 0 1 0 0 1 1 0 0 1 0 0 0 1 1 1 0 1 0 1 0 0 1 1 1 0 1 0 1 1 0 1 1]
# Expected: [1 1 1 1 0 1 0 1 0 0 0 0 1 0 1 1 1 1 0 1 0 0 0 1 1 0 1 0 0 1 0 0 0 1 0 1 1
#  1 0 1 0 1 1 0 0 0 1 0 0 1 0 1 0 0 1 1 0 0 0 1 0 0 1 1 1 0 0 1 1 1 1 0 1 0
#  0 0 1 1 1 0 1 0 1 1 0 0 0 0 1 0 1 0 1 0 1 1 1 1 0 1 1 0 0 1 0 0 1 1 1 0 1
#  1 0 1 0 0 0 1 1 1 0 1 1 1 0 1 0 0 0 0 1 1 0 1 1 0 1 1 0 1 1 0 1 0 0 0 0 1
#  1 1 1 1 0 1 1 1 0 0 0 1 0 1 0 0 0 1 1 1 1 1 1 1 1 0 0 1 0 0 0 0 1 0 0 0 1
#  1 1 0 1 1 1 1 0 0 0 1 1 0 1 1 1 0 0 1 1 0 1 1 1 0 1 1 1 1 1 0 1 0 1 1 0 0
#  0 0 1 1 1 0 1 1 0 1 1 1 1 1 1 0 0 1 1 1 1 0 1 1 1 0 1 1 1 0 1 1 0 1 1 1 0
#  0 1 1 0 1 0 1 0 0 0 0 1 0 1 0 1 1 1 1 0 1 0 1 0 1 1 0 1 1 1 1 1 0 1 0 1 0
#  1 1 1 1 0 1 0 1 1 1 1 1 0 1 0 0 1 0 0 1 1 1 1 1 0 0 0 1 1 0 1 0 1 1 1 1 1
#  0 1 1 0 1 0 1 0 1 1 0 1 0 1 0 0 0 1 1 1 1 0 0 1 1 1 0 1 0 0 1 1 0 0 1 0 0
#  1 1 1 1 0 0 1 1 1 1 1 1 0 1 1 1 1 0 0 0 1 0 1 0 0 1 0 1 1 1 1 1 1 1 0 0 1
#  1 0 1 1 1 1 1 0 0 1 1 0 1 1 0 1 0 0 1 1 0 1 1 1 1 1 0 1 0 1 1 1 1 1 1 0 0
#  0 1 0 1 0 1 0 0 1 1 0 1 0 1 1 0 1 1 1 1 1 1 0 0 0 0 1 1 1 1 1 0 1 0 0 1 1
#  0 1 1 0 1 0 0 1 1 1 1 0 0 0 0 1 1 0 1 0 0 0 0 0 0 1 1 1 1 0 1 1 0 1 0 1 1
#  1 1 1 0 1 1 0 1 0 1 0 1 1 0 1 0 1 1 1 0 0 1 0 0 1 1 1 1 1 1 1 1 1 1 1 0 0
#  0 0 0 0 0 1 0 0 1 1 0 0 1 0 0 0 1 1 1 0 1 0 1 0 0 1 1 1 0 1 0 1 0 0 1 1]

correct = accuracy_score(y_test_arg,predict_classes)
print("Accuracy: {}".format(correct))

#Accuracy: 0.9712351945854484