ML Python 100 - Impute Categorical with Dictionary Vectorizer

<p>KNIME Python Script: Convert Categorical Features with DictVectorizer and Store Metadata</p><p>----</p><p>Short Summary</p><p>This script prepares a KNIME input table for machine learning by&nbsp;<strong>encoding categorical variables into numerical features</strong>&nbsp;using scikit-learn’s&nbsp;DictVectorizer.</p><p>Steps it performs:</p><ol><li><p><strong>Load data</strong>&nbsp;from KNIME and reset the index to preserve row order.</p></li><li><p><strong>Identify feature groups</strong>: excluded columns, label column(s), numeric columns, categorical columns, and the rest.</p></li><li><p><strong>Vectorize categorical features</strong>:</p><ul><li><p>Convert categories to strings,</p></li><li><p>Apply&nbsp;DictVectorizer&nbsp;to one-hot encode them,</p></li><li><p>Combine the encoded categorical data with the rest of the dataset.</p></li></ul></li><li><p><strong>Generate metadata</strong>: Store lists of excluded, label, numeric, categorical, and transformed column names.</p></li><li><p><strong>Save the vectorizer vocabulary and feature names as JSON files</strong>&nbsp;for later use.</p></li><li><p><strong>Output results to KNIME</strong>:</p><ul><li><p>A table with metadata,</p></li><li><p>The transformed training dataset,</p></li><li><p>The trained vectorizer object,</p></li><li><p>The metadata dictionary.</p></li></ul></li></ol><p>👉 In essence, this script&nbsp;<strong>transforms categorical columns into a machine-learning–ready numeric format (one-hot encoding), saves the mapping, and outputs both data and metadata back into KNIME</strong>.</p>