Wine recognition dataset (tf.keras) ワイン分類

I understand that no need to use keras for this wine recognition. Scikit-learn is better, easy and good result. This is just less than 200 dataset. Just I try to use keras for small dataset of categorical analysis.

# Download dataset
import numpy as np
from sklearn.datasets import load_wine
wine = load_wine()

print(wine.DESCR)

.. _wine_dataset:

Wine recognition dataset
------------------------

**Data Set Characteristics:**

    :Number of Instances: 178 (50 in each of three classes)
    :Number of Attributes: 13 numeric, predictive attributes and the class
    :Attribute Information:
 		- Alcohol
 		- Malic acid
 		- Ash
		- Alcalinity of ash  
 		- Magnesium
		- Total phenols
 		- Flavanoids
 		- Nonflavanoid phenols
 		- Proanthocyanins
		- Color intensity
 		- Hue
 		- OD280/OD315 of diluted wines
 		- Proline

    - class:
            - class_0
            - class_1
            - class_2
		
    :Summary Statistics:
    
    ============================= ==== ===== ======= =====
                                   Min   Max   Mean     SD
    ============================= ==== ===== ======= =====
    Alcohol:                      11.0  14.8    13.0   0.8
    Malic Acid:                   0.74  5.80    2.34  1.12
    Ash:                          1.36  3.23    2.36  0.27
    Alcalinity of Ash:            10.6  30.0    19.5   3.3
    Magnesium:                    70.0 162.0    99.7  14.3
    Total Phenols:                0.98  3.88    2.29  0.63
    Flavanoids:                   0.34  5.08    2.03  1.00
    Nonflavanoid Phenols:         0.13  0.66    0.36  0.12
    Proanthocyanins:              0.41  3.58    1.59  0.57
    Colour Intensity:              1.3  13.0     5.1   2.3
    Hue:                          0.48  1.71    0.96  0.23
    OD280/OD315 of diluted wines: 1.27  4.00    2.61  0.71
    Proline:                       278  1680     746   315
    ============================= ==== ===== ======= =====

    :Missing Attribute Values: None
    :Class Distribution: class_0 (59), class_1 (71), class_2 (48)
    :Creator: R.A. Fisher
    :Donor: Michael Marshall (MARSHALL%PLU@io.arc.nasa.gov)
    :Date: July, 1988

This is a copy of UCI ML Wine recognition datasets.
https://archive.ics.uci.edu/ml/machine-learning-databases/wine/wine.data

The data is the results of a chemical analysis of wines grown in the same
region in Italy by three different cultivators. There are thirteen different
measurements taken for different constituents found in the three types of
wine.

Original Owners: 

Forina, M. et al, PARVUS - 
An Extendible Package for Data Exploration, Classification and Correlation. 
Institute of Pharmaceutical and Food Analysis and Technologies,
Via Brigata Salerno, 16147 Genoa, Italy.

Citation:

Lichman, M. (2013). UCI Machine Learning Repository
[https://archive.ics.uci.edu/ml]. Irvine, CA: University of California,
School of Information and Computer Science. 

.. topic:: References

  (1) S. Aeberhard, D. Coomans and O. de Vel, 
  Comparison of Classifiers in High Dimensional Settings, 
  Tech. Rep. no. 92-02, (1992), Dept. of Computer Science and Dept. of  
  Mathematics and Statistics, James Cook University of North Queensland. 
  (Also submitted to Technometrics). 

  The data was used with many others for comparing various 
  classifiers. The classes are separable, though only RDA 
  has achieved 100% correct classification. 
  (RDA : 100%, QDA 99.4%, LDA 98.9%, 1NN 96.1% (z-transformed data)) 
  (All results using the leave-one-out technique) 

  (2) S. Aeberhard, D. Coomans and O. de Vel, 
  "THE CLASSIFICATION PERFORMANCE OF RDA" 
  Tech. Rep. no. 92-01, (1992), Dept. of Computer Science and Dept. of 
  Mathematics and Statistics, James Cook University of North Queensland. 
  (Also submitted to Journal of Chemometrics).

wine.data

array([[1.423e+01, 1.710e+00, 2.430e+00, ..., 1.040e+00, 3.920e+00,
        1.065e+03],
       [1.320e+01, 1.780e+00, 2.140e+00, ..., 1.050e+00, 3.400e+00,
        1.050e+03],
       [1.316e+01, 2.360e+00, 2.670e+00, ..., 1.030e+00, 3.170e+00,
        1.185e+03],
       ...,
       [1.327e+01, 4.280e+00, 2.260e+00, ..., 5.900e-01, 1.560e+00,
        8.350e+02],
       [1.317e+01, 2.590e+00, 2.370e+00, ..., 6.000e-01, 1.620e+00,
        8.400e+02],
       [1.413e+01, 4.100e+00, 2.740e+00, ..., 6.100e-01, 1.600e+00,
        5.600e+02]])

wine.target

array([0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
       0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1,
       1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
       1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
       1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2,
       2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
       2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
       2, 2])

from sklearn.model_selection import train_test_split as split
x_train, x_test, y_train, y_test = split(wine.data,wine.target,train_size=0.8,test_size=0.2)

from __future__ import absolute_import, division, print_function, unicode_literals

# Install TensorFlow
try:
  # %tensorflow_version only exists in Colab.
  %tensorflow_version 2.x
except Exception:
  pass
  
import tensorflow as tf

# Create sequential model(Maybe no need to be build deep like this. More simple model better)
model = tf.keras.Sequential([
    tf.keras.layers.Dense(26, activation='relu', input_shape=(13,)),
    tf.keras.layers.Dense(52, activation='relu', input_shape=(26,)),
    tf.keras.layers.Dense(104, activation='relu', input_shape=(52,)),
    tf.keras.layers.Dense(3, activation='softmax')
])

model.summary()

Model: "sequential"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense (Dense)                (None, 26)                364       
_________________________________________________________________
dense_1 (Dense)              (None, 52)                1404      
_________________________________________________________________
dense_2 (Dense)              (None, 104)               5512      
_________________________________________________________________
dense_3 (Dense)              (None, 3)                 315       
=================================================================
Total params: 7,595
Trainable params: 7,595
Non-trainable params: 0
_________________________________________________________________

model.compile(loss='sparse_categorical_crossentropy', optimizer='adam', metrics=['acc'])

history = model.fit(x_train, y_train, batch_size=128, epochs=100, verbose=1)

Epoch 1/100
2/2 [==============================] - 0s 3ms/step - loss: 8.1767 - acc: 0.3451
Epoch 2/100
2/2 [==============================] - 0s 1ms/step - loss: 18.4229 - acc: 0.2817
Epoch 3/100
2/2 [==============================] - 0s 2ms/step - loss: 1.6280 - acc: 0.5000
Epoch 4/100
2/2 [==============================] - 0s 1ms/step - loss: 6.7892 - acc: 0.6127
Epoch 5/100
2/2 [==============================] - 0s 1ms/step - loss: 8.2274 - acc: 0.5775
Epoch 6/100
2/2 [==============================] - 0s 2ms/step - loss: 6.6641 - acc: 0.5141
Epoch 7/100
2/2 [==============================] - 0s 2ms/step - loss: 2.4120 - acc: 0.6479
Epoch 8/100
2/2 [==============================] - 0s 2ms/step - loss: 7.1879 - acc: 0.2817
Epoch 9/100
2/2 [==============================] - 0s 2ms/step - loss: 5.3963 - acc: 0.3169
Epoch 10/100
2/2 [==============================] - 0s 2ms/step - loss: 2.7827 - acc: 0.5845
Epoch 11/100
2/2 [==============================] - 0s 2ms/step - loss: 4.6227 - acc: 0.5352
Epoch 12/100
2/2 [==============================] - 0s 2ms/step - loss: 4.5851 - acc: 0.5070
Epoch 13/100
2/2 [==============================] - 0s 2ms/step - loss: 2.3347 - acc: 0.6479
Epoch 14/100
2/2 [==============================] - 0s 2ms/step - loss: 2.0858 - acc: 0.4366
Epoch 15/100
2/2 [==============================] - 0s 1ms/step - loss: 1.3559 - acc: 0.3662
Epoch 16/100
2/2 [==============================] - 0s 3ms/step - loss: 1.1846 - acc: 0.5000
Epoch 17/100
2/2 [==============================] - 0s 2ms/step - loss: 1.5474 - acc: 0.4437
Epoch 18/100
2/2 [==============================] - 0s 2ms/step - loss: 1.4257 - acc: 0.4366
Epoch 19/100
2/2 [==============================] - 0s 2ms/step - loss: 1.3411 - acc: 0.5282
Epoch 20/100
2/2 [==============================] - 0s 2ms/step - loss: 1.1221 - acc: 0.5282
Epoch 21/100
2/2 [==============================] - 0s 2ms/step - loss: 1.5945 - acc: 0.6620
Epoch 22/100
2/2 [==============================] - 0s 2ms/step - loss: 1.3009 - acc: 0.6197
Epoch 23/100
2/2 [==============================] - 0s 2ms/step - loss: 1.3326 - acc: 0.5282
Epoch 24/100
2/2 [==============================] - 0s 2ms/step - loss: 1.0436 - acc: 0.5845
Epoch 25/100
2/2 [==============================] - 0s 2ms/step - loss: 0.6736 - acc: 0.5986
Epoch 26/100
2/2 [==============================] - 0s 2ms/step - loss: 1.0627 - acc: 0.5141
Epoch 27/100
2/2 [==============================] - 0s 2ms/step - loss: 0.7542 - acc: 0.6408
Epoch 28/100
2/2 [==============================] - 0s 2ms/step - loss: 0.8888 - acc: 0.6338
Epoch 29/100
2/2 [==============================] - 0s 2ms/step - loss: 1.0726 - acc: 0.5352
Epoch 30/100
2/2 [==============================] - 0s 2ms/step - loss: 0.8842 - acc: 0.6690
Epoch 31/100
2/2 [==============================] - 0s 2ms/step - loss: 0.7021 - acc: 0.6338
Epoch 32/100
2/2 [==============================] - 0s 2ms/step - loss: 1.0238 - acc: 0.5563
Epoch 33/100
2/2 [==============================] - 0s 2ms/step - loss: 0.7450 - acc: 0.6338
Epoch 34/100
2/2 [==============================] - 0s 2ms/step - loss: 0.6461 - acc: 0.7042
Epoch 35/100
2/2 [==============================] - 0s 2ms/step - loss: 0.5955 - acc: 0.6831
Epoch 36/100
2/2 [==============================] - 0s 2ms/step - loss: 0.7565 - acc: 0.6268
Epoch 37/100
2/2 [==============================] - 0s 2ms/step - loss: 0.9491 - acc: 0.5704
Epoch 38/100
2/2 [==============================] - 0s 2ms/step - loss: 0.8987 - acc: 0.6408
Epoch 39/100
2/2 [==============================] - 0s 2ms/step - loss: 0.8005 - acc: 0.6268
Epoch 40/100
2/2 [==============================] - 0s 2ms/step - loss: 1.0723 - acc: 0.5634
Epoch 41/100
2/2 [==============================] - 0s 2ms/step - loss: 0.6473 - acc: 0.6761
Epoch 42/100
2/2 [==============================] - 0s 2ms/step - loss: 0.7843 - acc: 0.6479
Epoch 43/100
2/2 [==============================] - 0s 2ms/step - loss: 1.3531 - acc: 0.4859
Epoch 44/100
2/2 [==============================] - 0s 2ms/step - loss: 1.0187 - acc: 0.6408
Epoch 45/100
2/2 [==============================] - 0s 3ms/step - loss: 1.0043 - acc: 0.6479
Epoch 46/100
2/2 [==============================] - 0s 2ms/step - loss: 1.0806 - acc: 0.5915
Epoch 47/100
2/2 [==============================] - 0s 2ms/step - loss: 0.5988 - acc: 0.7183
Epoch 48/100
2/2 [==============================] - 0s 2ms/step - loss: 0.5754 - acc: 0.6831
Epoch 49/100
2/2 [==============================] - 0s 2ms/step - loss: 0.9460 - acc: 0.6338
Epoch 50/100
2/2 [==============================] - 0s 2ms/step - loss: 0.7031 - acc: 0.6479
Epoch 51/100
2/2 [==============================] - 0s 2ms/step - loss: 0.6227 - acc: 0.7394
Epoch 52/100
2/2 [==============================] - 0s 5ms/step - loss: 0.6894 - acc: 0.6831
Epoch 53/100
2/2 [==============================] - 0s 2ms/step - loss: 0.7101 - acc: 0.6620
Epoch 54/100
2/2 [==============================] - 0s 2ms/step - loss: 0.7836 - acc: 0.6690
Epoch 55/100
2/2 [==============================] - 0s 2ms/step - loss: 0.6260 - acc: 0.6831
Epoch 56/100
2/2 [==============================] - 0s 2ms/step - loss: 0.6250 - acc: 0.6972
Epoch 57/100
2/2 [==============================] - 0s 2ms/step - loss: 0.5411 - acc: 0.7254
Epoch 58/100
2/2 [==============================] - 0s 2ms/step - loss: 0.7616 - acc: 0.6901
Epoch 59/100
2/2 [==============================] - 0s 2ms/step - loss: 0.9814 - acc: 0.5986
Epoch 60/100
2/2 [==============================] - 0s 2ms/step - loss: 0.5598 - acc: 0.7042
Epoch 61/100
2/2 [==============================] - 0s 2ms/step - loss: 1.1271 - acc: 0.6620
Epoch 62/100
2/2 [==============================] - 0s 2ms/step - loss: 0.6484 - acc: 0.7254
Epoch 63/100
2/2 [==============================] - 0s 2ms/step - loss: 0.6730 - acc: 0.6690
Epoch 64/100
2/2 [==============================] - 0s 2ms/step - loss: 0.5882 - acc: 0.7113
Epoch 65/100
2/2 [==============================] - 0s 2ms/step - loss: 1.2114 - acc: 0.5986
Epoch 66/100
2/2 [==============================] - 0s 2ms/step - loss: 0.7267 - acc: 0.6479
Epoch 67/100
2/2 [==============================] - 0s 2ms/step - loss: 0.8106 - acc: 0.6972
Epoch 68/100
2/2 [==============================] - 0s 2ms/step - loss: 1.5999 - acc: 0.4859
Epoch 69/100
2/2 [==============================] - 0s 2ms/step - loss: 1.1265 - acc: 0.6338
Epoch 70/100
2/2 [==============================] - 0s 2ms/step - loss: 1.2845 - acc: 0.6268
Epoch 71/100
2/2 [==============================] - 0s 2ms/step - loss: 0.7232 - acc: 0.6761
Epoch 72/100
2/2 [==============================] - 0s 2ms/step - loss: 1.4645 - acc: 0.5634
Epoch 73/100
2/2 [==============================] - 0s 2ms/step - loss: 0.7787 - acc: 0.6690
Epoch 74/100
2/2 [==============================] - 0s 2ms/step - loss: 0.7638 - acc: 0.6690
Epoch 75/100
2/2 [==============================] - 0s 2ms/step - loss: 0.8269 - acc: 0.6761
Epoch 76/100
2/2 [==============================] - 0s 2ms/step - loss: 0.8411 - acc: 0.6620
Epoch 77/100
2/2 [==============================] - 0s 2ms/step - loss: 0.6552 - acc: 0.6972
Epoch 78/100
2/2 [==============================] - 0s 2ms/step - loss: 0.5763 - acc: 0.7113
Epoch 79/100
2/2 [==============================] - 0s 2ms/step - loss: 0.9468 - acc: 0.6338
Epoch 80/100
2/2 [==============================] - 0s 2ms/step - loss: 0.7569 - acc: 0.6479
Epoch 81/100
2/2 [==============================] - 0s 2ms/step - loss: 0.6522 - acc: 0.6901
Epoch 82/100
2/2 [==============================] - 0s 2ms/step - loss: 0.5492 - acc: 0.7254
Epoch 83/100
2/2 [==============================] - 0s 2ms/step - loss: 0.6368 - acc: 0.6901
Epoch 84/100
2/2 [==============================] - 0s 2ms/step - loss: 0.5285 - acc: 0.7113
Epoch 85/100
2/2 [==============================] - 0s 2ms/step - loss: 0.7157 - acc: 0.7042
Epoch 86/100
2/2 [==============================] - 0s 2ms/step - loss: 0.5902 - acc: 0.7183
Epoch 87/100
2/2 [==============================] - 0s 2ms/step - loss: 0.6237 - acc: 0.7324
Epoch 88/100
2/2 [==============================] - 0s 2ms/step - loss: 0.7108 - acc: 0.7113
Epoch 89/100
2/2 [==============================] - 0s 2ms/step - loss: 0.6581 - acc: 0.6690
Epoch 90/100
2/2 [==============================] - 0s 2ms/step - loss: 0.8553 - acc: 0.6479
Epoch 91/100
2/2 [==============================] - 0s 2ms/step - loss: 0.6352 - acc: 0.7042
Epoch 92/100
2/2 [==============================] - 0s 2ms/step - loss: 0.5463 - acc: 0.7465
Epoch 93/100
2/2 [==============================] - 0s 2ms/step - loss: 0.7264 - acc: 0.7042
Epoch 94/100
2/2 [==============================] - 0s 2ms/step - loss: 1.1365 - acc: 0.6549
Epoch 95/100
2/2 [==============================] - 0s 2ms/step - loss: 0.5548 - acc: 0.7254
Epoch 96/100
2/2 [==============================] - 0s 2ms/step - loss: 1.1269 - acc: 0.5915
Epoch 97/100
2/2 [==============================] - 0s 2ms/step - loss: 0.9596 - acc: 0.6268
Epoch 98/100
2/2 [==============================] - 0s 2ms/step - loss: 0.5778 - acc: 0.7465
Epoch 99/100
2/2 [==============================] - 0s 2ms/step - loss: 0.8838 - acc: 0.6268
Epoch 100/100
2/2 [==============================] - 0s 2ms/step - loss: 0.9017 - acc: 0.6690

# Evaluation
score = model.evaluate(x_test, y_test, batch_size = 32)
print("損失係数 loss : " + str(score[0]) + ", 正解率 accuracy : " + str(score[1]*100) + "% ")

# Prepare virtual test data and evaluate it.
import numpy as np
x = np.array([[11.23, 1.41, 3.43, 12.6, 167, 2.00, 4.06, 0.38, 2.30, 5.64, 1.04, 3.92, 1065]])
r = model.predict(x)
print(r)
answer = r.argmax()

if answer == 0:
  print("This wine is class 0.")
elif answer == 1:
  print("This wine is class 1.")
else:
  print("This wine is class 2.")

2/2 [==============================] - 0s 2ms/step - loss: 1.3145 - acc: 0.5556
損失係数 loss : 1.3145475387573242, 正解率 accuracy : 55.55555820465088% 
[[0.92950535 0.00265262 0.06784209]]
This wine is class 0.

Accuracy is not good score. If using sklearn, easy to find good result. Clearly there’s some overlearning going on. Seems good to use EarlyStopping. But this is just test. Let’s take a look at plot.

import matplotlib.pyplot as plt

# list all data in history
print(history.history.keys())

plt.plot(history.history['acc'])
plt.title('Accuracy')
plt.ylabel('accuracy')
plt.xlabel('epoch')
plt.legend(['train'], loc='upper left')
plt.show()

plt.plot(history.history['loss'])
plt.title('Model loss')
plt.ylabel('Loss')
plt.xlabel('Epoch')
plt.legend(['train', 'test'], loc='upper left')
plt.show()

# Create model using dropout
model = tf.keras.Sequential([
    tf.keras.layers.Dense(26, activation='relu', input_shape=(13,)),
    tf.keras.layers.Dropout(0.5),
    tf.keras.layers.Dense(52, activation='relu', input_shape=(26,)),
    tf.keras.layers.Dropout(0.5),
    tf.keras.layers.Dense(104, activation='relu', input_shape=(52,)),
    tf.keras.layers.Dense(3, activation='softmax')
])

model.summary()

Model: "sequential_1"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense_4 (Dense)              (None, 26)                364       
_________________________________________________________________
dropout (Dropout)            (None, 26)                0         
_________________________________________________________________
dense_5 (Dense)              (None, 52)                1404      
_________________________________________________________________
dropout_1 (Dropout)          (None, 52)                0         
_________________________________________________________________
dense_6 (Dense)              (None, 104)               5512      
_________________________________________________________________
dense_7 (Dense)              (None, 3)                 315       
=================================================================
Total params: 7,595
Trainable params: 7,595
Non-trainable params: 0
_________________________________________________________________

model.compile(loss='sparse_categorical_crossentropy', optimizer='adam', metrics=['acc'])

history = model.fit(x_train, y_train, batch_size=128, epochs=100, verbose=1)

Epoch 1/100
2/2 [==============================] - 0s 2ms/step - loss: 78.4047 - acc: 0.3732
Epoch 2/100
2/2 [==============================] - 0s 3ms/step - loss: 72.1076 - acc: 0.3944
Epoch 3/100
2/2 [==============================] - 0s 2ms/step - loss: 50.2278 - acc: 0.3803
Epoch 4/100
2/2 [==============================] - 0s 2ms/step - loss: 42.7940 - acc: 0.4225
Epoch 5/100
2/2 [==============================] - 0s 2ms/step - loss: 45.1331 - acc: 0.3310
Epoch 6/100
2/2 [==============================] - 0s 2ms/step - loss: 43.4943 - acc: 0.2746
Epoch 7/100
2/2 [==============================] - 0s 2ms/step - loss: 47.0344 - acc: 0.2958
Epoch 8/100
2/2 [==============================] - 0s 2ms/step - loss: 34.0279 - acc: 0.3451
Epoch 9/100
2/2 [==============================] - 0s 2ms/step - loss: 32.5107 - acc: 0.3592
Epoch 10/100
2/2 [==============================] - 0s 2ms/step - loss: 28.6999 - acc: 0.3732
Epoch 11/100
2/2 [==============================] - 0s 2ms/step - loss: 40.4330 - acc: 0.3028
Epoch 12/100
2/2 [==============================] - 0s 2ms/step - loss: 35.4584 - acc: 0.2676
Epoch 13/100
2/2 [==============================] - 0s 2ms/step - loss: 34.7127 - acc: 0.2676
Epoch 14/100
2/2 [==============================] - 0s 2ms/step - loss: 24.5392 - acc: 0.3803
Epoch 15/100
2/2 [==============================] - 0s 3ms/step - loss: 28.7477 - acc: 0.3380
Epoch 16/100
2/2 [==============================] - 0s 2ms/step - loss: 29.9051 - acc: 0.3028
Epoch 17/100
2/2 [==============================] - 0s 2ms/step - loss: 23.9948 - acc: 0.3310
Epoch 18/100
2/2 [==============================] - 0s 2ms/step - loss: 22.5187 - acc: 0.4296
Epoch 19/100
2/2 [==============================] - 0s 2ms/step - loss: 22.2603 - acc: 0.4225
Epoch 20/100
2/2 [==============================] - 0s 2ms/step - loss: 21.4563 - acc: 0.3521
Epoch 21/100
2/2 [==============================] - 0s 2ms/step - loss: 22.8434 - acc: 0.3592
Epoch 22/100
2/2 [==============================] - 0s 2ms/step - loss: 21.4928 - acc: 0.2817
Epoch 23/100
2/2 [==============================] - 0s 2ms/step - loss: 19.7906 - acc: 0.3944
Epoch 24/100
2/2 [==============================] - 0s 2ms/step - loss: 16.5452 - acc: 0.4296
Epoch 25/100
2/2 [==============================] - 0s 2ms/step - loss: 20.4858 - acc: 0.3310
Epoch 26/100
2/2 [==============================] - 0s 2ms/step - loss: 19.8785 - acc: 0.3169
Epoch 27/100
2/2 [==============================] - 0s 2ms/step - loss: 18.2515 - acc: 0.3169
Epoch 28/100
2/2 [==============================] - 0s 3ms/step - loss: 19.0335 - acc: 0.3521
Epoch 29/100
2/2 [==============================] - 0s 2ms/step - loss: 19.0035 - acc: 0.3239
Epoch 30/100
2/2 [==============================] - 0s 2ms/step - loss: 18.0972 - acc: 0.3239
Epoch 31/100
2/2 [==============================] - 0s 2ms/step - loss: 21.0755 - acc: 0.2746
Epoch 32/100
2/2 [==============================] - 0s 2ms/step - loss: 14.1798 - acc: 0.4296
Epoch 33/100
2/2 [==============================] - 0s 2ms/step - loss: 15.9207 - acc: 0.3310
Epoch 34/100
2/2 [==============================] - 0s 1ms/step - loss: 16.9278 - acc: 0.3662
Epoch 35/100
2/2 [==============================] - 0s 1ms/step - loss: 15.2984 - acc: 0.3873
Epoch 36/100
2/2 [==============================] - 0s 2ms/step - loss: 15.5278 - acc: 0.3521
Epoch 37/100
2/2 [==============================] - 0s 2ms/step - loss: 18.0193 - acc: 0.3732
Epoch 38/100
2/2 [==============================] - 0s 1ms/step - loss: 15.0588 - acc: 0.3521
Epoch 39/100
2/2 [==============================] - 0s 2ms/step - loss: 14.2689 - acc: 0.3662
Epoch 40/100
2/2 [==============================] - 0s 2ms/step - loss: 13.7116 - acc: 0.3521
Epoch 41/100
2/2 [==============================] - 0s 2ms/step - loss: 13.9965 - acc: 0.3169
Epoch 42/100
2/2 [==============================] - 0s 2ms/step - loss: 15.1268 - acc: 0.3028
Epoch 43/100
2/2 [==============================] - 0s 3ms/step - loss: 14.5549 - acc: 0.3169
Epoch 44/100
2/2 [==============================] - 0s 2ms/step - loss: 11.1286 - acc: 0.3944
Epoch 45/100
2/2 [==============================] - 0s 2ms/step - loss: 16.3549 - acc: 0.3380
Epoch 46/100
2/2 [==============================] - 0s 2ms/step - loss: 13.5113 - acc: 0.3310
Epoch 47/100
2/2 [==============================] - 0s 2ms/step - loss: 9.9733 - acc: 0.3662
Epoch 48/100
2/2 [==============================] - 0s 2ms/step - loss: 12.6871 - acc: 0.3732
Epoch 49/100
2/2 [==============================] - 0s 2ms/step - loss: 11.7777 - acc: 0.4085
Epoch 50/100
2/2 [==============================] - 0s 2ms/step - loss: 13.7046 - acc: 0.3732
Epoch 51/100
2/2 [==============================] - 0s 2ms/step - loss: 11.4345 - acc: 0.3451
Epoch 52/100
2/2 [==============================] - 0s 3ms/step - loss: 11.6702 - acc: 0.3521
Epoch 53/100
2/2 [==============================] - 0s 3ms/step - loss: 10.4102 - acc: 0.4014
Epoch 54/100
2/2 [==============================] - 0s 2ms/step - loss: 10.1209 - acc: 0.3873
Epoch 55/100
2/2 [==============================] - 0s 2ms/step - loss: 11.5787 - acc: 0.3873
Epoch 56/100
2/2 [==============================] - 0s 2ms/step - loss: 13.3705 - acc: 0.3028
Epoch 57/100
2/2 [==============================] - 0s 2ms/step - loss: 9.7055 - acc: 0.3873
Epoch 58/100
2/2 [==============================] - 0s 2ms/step - loss: 11.2023 - acc: 0.3732
Epoch 59/100
2/2 [==============================] - 0s 2ms/step - loss: 11.5734 - acc: 0.4085
Epoch 60/100
2/2 [==============================] - 0s 2ms/step - loss: 10.6941 - acc: 0.3592
Epoch 61/100
2/2 [==============================] - 0s 2ms/step - loss: 10.2570 - acc: 0.4014
Epoch 62/100
2/2 [==============================] - 0s 2ms/step - loss: 11.8426 - acc: 0.3310
Epoch 63/100
2/2 [==============================] - 0s 2ms/step - loss: 10.0226 - acc: 0.3521
Epoch 64/100
2/2 [==============================] - 0s 2ms/step - loss: 9.2977 - acc: 0.3521
Epoch 65/100
2/2 [==============================] - 0s 2ms/step - loss: 9.4539 - acc: 0.4085
Epoch 66/100
2/2 [==============================] - 0s 2ms/step - loss: 10.3128 - acc: 0.3521
Epoch 67/100
2/2 [==============================] - 0s 2ms/step - loss: 8.8057 - acc: 0.4155
Epoch 68/100
2/2 [==============================] - 0s 2ms/step - loss: 12.8116 - acc: 0.2958
Epoch 69/100
2/2 [==============================] - 0s 2ms/step - loss: 10.6053 - acc: 0.3521
Epoch 70/100
2/2 [==============================] - 0s 2ms/step - loss: 7.3333 - acc: 0.4225
Epoch 71/100
2/2 [==============================] - 0s 3ms/step - loss: 8.6286 - acc: 0.4296
Epoch 72/100
2/2 [==============================] - 0s 3ms/step - loss: 9.6019 - acc: 0.3451
Epoch 73/100
2/2 [==============================] - 0s 2ms/step - loss: 9.4916 - acc: 0.3732
Epoch 74/100
2/2 [==============================] - 0s 2ms/step - loss: 10.0909 - acc: 0.4014
Epoch 75/100
2/2 [==============================] - 0s 2ms/step - loss: 10.7403 - acc: 0.2958
Epoch 76/100
2/2 [==============================] - 0s 3ms/step - loss: 8.9081 - acc: 0.3451
Epoch 77/100
2/2 [==============================] - 0s 2ms/step - loss: 8.3458 - acc: 0.3662
Epoch 78/100
2/2 [==============================] - 0s 2ms/step - loss: 8.5031 - acc: 0.2676
Epoch 79/100
2/2 [==============================] - 0s 3ms/step - loss: 7.7347 - acc: 0.3944
Epoch 80/100
2/2 [==============================] - 0s 2ms/step - loss: 7.4269 - acc: 0.3592
Epoch 81/100
2/2 [==============================] - 0s 2ms/step - loss: 9.6820 - acc: 0.3239
Epoch 82/100
2/2 [==============================] - 0s 2ms/step - loss: 8.8056 - acc: 0.3451
Epoch 83/100
2/2 [==============================] - 0s 2ms/step - loss: 8.6425 - acc: 0.3380
Epoch 84/100
2/2 [==============================] - 0s 2ms/step - loss: 7.6924 - acc: 0.3380
Epoch 85/100
2/2 [==============================] - 0s 2ms/step - loss: 6.8624 - acc: 0.4437
Epoch 86/100
2/2 [==============================] - 0s 2ms/step - loss: 8.5389 - acc: 0.3380
Epoch 87/100
2/2 [==============================] - 0s 2ms/step - loss: 8.0148 - acc: 0.3803
Epoch 88/100
2/2 [==============================] - 0s 2ms/step - loss: 9.7312 - acc: 0.3592
Epoch 89/100
2/2 [==============================] - 0s 2ms/step - loss: 6.2159 - acc: 0.3803
Epoch 90/100
2/2 [==============================] - 0s 2ms/step - loss: 6.4581 - acc: 0.3873
Epoch 91/100
2/2 [==============================] - 0s 2ms/step - loss: 6.3186 - acc: 0.3732
Epoch 92/100
2/2 [==============================] - 0s 2ms/step - loss: 6.8777 - acc: 0.4155
Epoch 93/100
2/2 [==============================] - 0s 2ms/step - loss: 6.2091 - acc: 0.4014
Epoch 94/100
2/2 [==============================] - 0s 2ms/step - loss: 7.0392 - acc: 0.3592
Epoch 95/100
2/2 [==============================] - 0s 2ms/step - loss: 6.7406 - acc: 0.3451
Epoch 96/100
2/2 [==============================] - 0s 2ms/step - loss: 6.7102 - acc: 0.4225
Epoch 97/100
2/2 [==============================] - 0s 2ms/step - loss: 6.9462 - acc: 0.3239
Epoch 98/100
2/2 [==============================] - 0s 2ms/step - loss: 5.5881 - acc: 0.4296
Epoch 99/100
2/2 [==============================] - 0s 2ms/step - loss: 6.5911 - acc: 0.3592
Epoch 100/100
2/2 [==============================] - 0s 2ms/step - loss: 7.3447 - acc: 0.4225

And it got even worse. As It is expected, no need to use dropout.. I think it is better to use scikit-learn for such small data. Thanks.