tensorflow-kfold-cross-validation.py

import tensorflow as tf
import pandas as pd
from sklearn.model_selection import StratifiedKFold


def tf_cross_val_score(X, y, x_col_nms, n_classes):
    """
    Takes features, number of classes, labels and feature vector names and returns 3 validation scores.
    
    *X* Numpy ndarray - features
    *y* Numpy ndarray - labels
    *x_col_nms* a list() - feature dataset column names
    *n_classes* int - number of classes
    """
    scores = []
    kfold = StratifiedKFold(n_splits = 3, random_state = 42)
    for train_index, test_index in kfold.split(X, y):
        X_train_df, X_test_df = pd.DataFrame(data = X[train_index], columns = x_col_nms),\
            pd.DataFrame(data = X[test_index], columns = x_col_nms)
        y_train, y_test = y[train_index], y[test_index]
        
        clfr = tf.estimator.DNNClassifier(
            feature_columns = columns_feat,
            hidden_units=[10, 10], n_classes = n_classes)
        clfr.train(input_fn = lambda : train_function(X_train_df, y_train, 1000), steps = 100)
        scores.append(clfr.evaluate(input_fn = lambda : evaluation_function(X_test_df, y_test, 1000))['accuracy'])
        
    return pd.np.array(scores)