This a demo on Tensorflow Feature Column

In this demo a represtation of tf.feature_column is presented based on well known Titanic dataset, which available on

List of contents

1. Visulizing the data set as a DataFrame
2. Presenting a demo on Featute_Column

Visulizing the data set as a DataFrame

The first part of this demo focuses on dataset it self and gives some data visualization based on pandas.dataframe, which shows the main charectristics of the data itself.

Presenting a demo on Featute_Column

The second part, tries to cover some main tf.feature_columns including BucketizedColumn, NumericColumn, CategoricalColumnWithVocabulary, and wrapping CaltegoricalColumn using IndicatorColumn.
A CrossFeatureColumn is also included to demonstrate how individual features can mix together in order to privide a more depth knowledge of correlational features such as age and sex. Tools like FACET and Embeding Projector are out there on the internet to monitor the effecr of single individual featuter to an other one, and eventually identify the correlation between two individual features.
Moreover, the differences between the raw data and DenseFeatures is presented in the output to demonstrate how feature_column transform raw data in a form that is suitable for a tf.model