import numpy as np
import random
import pandas as pd
import keras
from keras.models import Sequential
from keras.layers import Dense
from sklearn.cross_validation import train_test_split

def gen_variance_dataset(n_elements, n_features):
    n_rows = int(n_elements / n_features)
    data_matrix = np.random.randint(1, 100, n_rows * n_features).reshape((n_rows, n_features))
    data = pd.DataFrame(data_matrix, columns=["feature_" + str(i) for i in range(n_features)])

    data["target"] = np.std(data_matrix, axis=1)

    return data

def build_model(n_features, n_hidden=10):

    model = Sequential()
    model.add(Dense(n_features, input_dim=n_features, kernel_initializer='normal', activation='relu'))
    for i in range(n_hidden):
        model.add(Dense(n_features,
                        activation="linear",
                        kernel_initializer="normal"))
    model.add(Dense(1, kernel_initializer='normal'))

    model.compile(loss='mean_squared_error', optimizer='adam')
    return model


data = gen_variance_dataset(100000, 5)
data.to_csv("variance_data.csv")
features = [col for col in data.columns if col != "target"]
model = build_model(len(features), 12)

X_train, X_test, y_train, y_test = train_test_split(data[features],
                                                    data["target"],
                                                    test_size=0.2)

model.fit(X_train.as_matrix(),
          y_train.as_matrix(),
          epochs=1000,
          batch_size=256,
          verbose=1)

evaluate = model.evaluate(X_test.as_matrix(),
                          y_test.as_matrix())

print("\n\nEvaluation {}".format(evaluate))