Indicator Columns
Learn about the indicator feature columns for the ML model's input layer.
We'll cover the following
Chapter Goals:
- Process the indicator feature columns used for the machine learning model’s input layer
A. One-hot indicators
The remaining non-numeric features in the dataset are the categorical features. Each of these features contains values that can be separated into a fixed number of distinct categories. For example, the 'IsHoliday' feature contains the categories 0 and 1, while the 'Type' feature contains the categories A, B, and C.
These two features specifically ('IsHoliday' and 'Type') will be the indicator features for the dataset. Since indicator features are categorical, this means they will be represented by one-hot vectors when aggregated into the model’s input layer.
B. Categorical column base
When creating the indicator feature columns, we start off with a categorical column base. The categorical column specifies the distinct categories of the feature, as well as the feature’s type. We’ll use the tf.feature_column.categorical_column_with_vocabulary_list list to create the categorical columns.
import tensorflow as tftype_col = tf.feature_column.categorical_column_with_vocabulary_list('Type', ['A', 'B', 'C'], dtype=tf.string)holiday_col = tf.feature_column.categorical_column_with_vocabulary_list('IsHoliday', [0, 1], dtype=tf.int64)
We can then convert the categorical column bases into indicator feature columns with the tf.feature_column.indicator_column function.
import tensorflow as tftype_feature_col = tf.feature_column.indicator_column(type_col)holiday_feature_col = tf.feature_column.indicator_column(holiday_col)
Time to Code!
In this chapter you’ll be creating the indicator feature columns for the dataset by completing the add_indicator_columns function. We’ve already filled the function with skeleton code that iterates through the indicator features in the dataset.
When creating the indicator feature columns, we need to specify the correct datatype. The 'Type' feature column will have datatype tf.string, while the 'IsHoliday' feature column will have datatype tf.int64.
Set dtype equal to the correct datatype, depending on what feature_name is.
Each indicator feature column is built from a vocabulary list. The vocabulary list comes from the unique values of the feature in the final_dataset DataFrame.
Set vocab_list equal to the unique values in final_dataset[feature_name], cast as a list.
Using the vocabulary list and datatype of the feature, we’ll create the categorical column for the feature.
Set vocab_col equal to tf.feature_column.categorical_column_with_vocabulary_list with feature_name and vocab_list as the required arguments, as well as dtype for the dtype keyword argument.
After creating the categorical column, we can convert it into the required indicator feature.
Set feature_col equal to tf.feature_column.indicator_column applied to vocab_col. Then append feature_col to the end of the feature_columns list.
import tensorflow as tf# Add the indicator feature columns to the list of feature columnsdef add_indicator_columns(final_dataset, feature_columns):indicator_features = ['IsHoliday', 'Type']for feature_name in indicator_features:# CODE HERE
Get hands-on with 1300+ tech skills courses.