A multiple-choice survey allows respondents to select multiple options per question. You must represent every respondent’s selections completely in a dataset to train a logistic regression model. Which approach provides a comprehensive representation suitable for logistic regression?

Creating one-hot (binary indicator) features is the most suitable approach because logistic regression models require numerical input. When respondents can select multiple options, a one-hot encoding transforms each possible option into a separate binary feature (0 or 1), indicating whether that option was selected. This method comprehensively captures all selections without losing information and provides the necessary numerical format for the model. Binning would group responses, losing the distinct information about each selected option. Amazon Mechanical Turk is a crowdsourcing service for labeling, not a data transformation method for this scenario. Amazon Textract is for extracting text from documents, not for converting survey responses into numerical features in this manner.