In this article, we will be using tf.estimator.LinearClassifier to build a model and train it on the famous titanic dataset. All of this will be done by using the TensorFlow API.
Importing Libraries
Python libraries make it easy for us to handle the data and perform typical and complex tasks with a single line of code.
- Pandas – This library helps to load the data frame in a 2D array format and has multiple functions to perform analysis tasks in one go.
- Numpy – Numpy arrays are very fast and can perform large computations in a very short time.
- Matplotlib/Seaborn – This library is used to draw visualizations.
Python3
import tensorflow as tf import tensorflow.feature_column as fc import numpy as np import pandas as pd import matplotlib.pyplot as plt import warnings warnings.filterwarnings('ignore') |
Importing Dataset
We will import the dataset by using the Tensorflow API for datasets and then load it into the panda’s data frame.
Python3
x_train = pd.read_csv( x_val = pd.read_csv( x_train.head() |
Output:
Python3
y_train = x_train.pop('survived') y_val = x_val.pop('survived') |
We will need the data for the categorical columns and the numeric(continuous) column present in the dataset separately to initialize our Linear Classifier model.
Python3
objects = [] numerics = [] for col in x_train.columns: if x_train[col].dtype == 'object': objects.append(col) elif x_train[col].dtype == 'int': objects.append(col) else: numerics.append(col) print(objects) print(numerics) |
Output:
['sex', 'n_siblings_spouses', 'parch', 'class', 'deck', 'embark_town', 'alone'] ['age', 'fare']
Python3
feat_cols = [] for feat_name in objects: vocabulary = x_train[feat_name].unique() feat_cols.append(fc.categorical_column_with_vocabulary_list(feat_name, vocabulary)) for feat_name in numerics: feat_cols.append(fc.numeric_column(feat_name, dtype=tf.float32)) |
We need to make a callable function that can be passed to the LinearClassifier function.
Python3
def make_input_fn(data, label, num_epochs=10, shuffle=True, batch_size=32): def input_function(): ds = tf.data.Dataset\ .from_tensor_slices((dict(data), label)) if shuffle: ds = ds.shuffle(1000) ds = ds.batch(batch_size)\ .repeat(num_epochs) return ds return input_function train_input_fn = make_input_fn(x_train, y_train) val_input_fn = make_input_fn(x_val, y_val, num_epochs=1, shuffle=False) |
Now we are good to go to train the tf.estimator.LinearClassifier model using the titanic dataset. Linear Classifier as the name suggests is a Linear model which is used to learn decision boundaries between multiple classes of the object but that should be Linear not non-Linear as we do so in the SVM algorithm.
LinearClassifier Model
Python3
linear_est = tf.estimator.LinearClassifier(feature_columns=feat_cols) linear_est.train(train_input_fn) result = linear_est.evaluate(val_input_fn) print(result) |
Output:
{'accuracy': 0.75,
'accuracy_baseline': 0.625,
'auc': 0.8377411,
'auc_precision_recall': 0.7833674,
'average_loss': 0.47364476,
'label/mean': 0.375, 'loss': 0.4666896,
'precision': 0.6666667,
'prediction/mean': 0.37083066,
'recall': 0.6666667,
'global_step': 200}
Here we can observe that the model has been evaluated on multiple matrices using the validation dataset and the accuracy obtained is also very satisfactory.
