An introduction to Keras-Classification.

Keras is a deep learning library written in Python which is running on top of TensorFlow, CNTK, or Theano.

Object

Learn a basic knowledge about the Keras
Learn classification

Preparing data
Keras Model
Building Model
Fit model
Run of codes

Preparing data

Here we consider the iris data that is well-known data set which often used for the classification, the following codes import data from from the UC Irvine Machine Learning Repository,

>>> import pandas as pd
>>> dataset = pd.read_csv("http://archive.ics.uci.edu/ml/machine-learning-databases/iris/iris.data",header=None,names=['SepalLength','SepalWidth','
PetalLength','PetalWidth','Species'])
>>> dataset.head()
   SepalLength  SepalWidth  PetalLength  PetalWidth      Species
0          5.1         3.5          1.4         0.2  Iris-setosa
1          4.9         3.0          1.4         0.2  Iris-setosa
2          4.7         3.2          1.3         0.2  Iris-setosa
3          4.6         3.1          1.5         0.2  Iris-setosa

4          5.0         3.6          1.4         0.2  Iris-setosa
>>> dataset.dtypes
SepalLength    float64
SepalWidth     float64
PetalLength    float64
PetalWidth     float64
Species         object
dtype: object

Define the dependent variable and covariates. The dependent variable (Species) is categorical and should change to the dummy variables, it can be done using simple codes:

X = dataset.values[:,0:4]
Y = pd.get_dummies(dataset['Species'])

Keras Model

The model part of neural networks includes the sequential layers so, the following code imports the sequential module which is a linear stack of layers and add it to model

from keras.models import Sequential
model = Sequential()

Building Model

You can also simply add layers using .add() , one needs to define the structure of layer. The simplest layer is the densely-connected NN layer,

from keras.layers import Dense

The following codes show a baseline neural network model for our data; first data line contains 10 neurons, since we have four variables, we assign input_dim=4. The second line creates output and emphasizes it has three values, one for each class.

model.add(Dense(10, input_dim=4, activation='relu'))
model.add(Dense(3, activation='softmax'))

Fitting Model

Once the model is built, it should be compiled, the following code define a loss function, optimizer, and a metric parameter.

model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])

The following functions use X (covariates) and Y (dependent variable), and select subsamples, batch_size, to train the model

model.fit(X, Y, nb_epoch=200, batch_size=10)
accuracy = model.evaluate(X,Y)
print("%s: %.2f%%" % (model.metrics_names[1], accuracy[1]*100))

References

There are very interesting websites that might be useful to
[Web1] https://machinelearningmastery.com/multi-class-classification-tutorial-keras-deep-learning-library/