How would you approach building an unsupervised learning model to detect outliers in a dataset with 6 behavioral variables and 25 contextual attributes, which needs normalization? Additionally, what evaluation metrics and visualizations would you use to measure the performance of the model?

Interview question asked to Data Scientists interviewing at Affirm, Plaid, Nubank and others: How would you approach building an unsupervised learning model to detect outliers in a dataset with 6 behavioral variables and 25 contextual attributes, which needs normalization? Additionally, what evaluation metrics and visualizations would you use to measure the performance of the model?.