Prepare Titanic data for model training

This 2019 Tech Talk used the Titanic dataset to show the work that happens before model training. The session covered ingestion, preparation, and feature engineering in Databricks with Python.

The archive preserves the session outline. It does not reproduce a current Databricks tutorial, and the original embedded slide player did not survive the site migration.

Why preparation took center stage

Even a small dataset can expose difficult questions. A field may be missing, a category may need a consistent definition, or a feature may only make sense with domain knowledge.

The talk used those decisions to show why data engineering and data science need to work together. Preparing the table is not a neutral step: every transformation changes what the model can learn.

Session path

1. Inspect the Titanic source fields and missing values.
2. Decide which fields could support the prediction task.
3. Create features in a repeatable notebook workflow.
4. Produce a training table for the next modeling session.

The original Slideshare reference was 150788201. Keep that ID with the archive if the deck is recovered.

The nine-part series

1. Data Science overview with Databricks
2. Titanic survival prediction with Azure Machine Learning Studio and Kaggle
3. Data engineering with the Titanic dataset, Databricks, and Python
4. Titanic with Databricks and Spark ML
5. Titanic with Databricks and Azure Machine Learning
6. Titanic with Databricks and AutoML
7. Titanic with Databricks and MLflow
8. Titanic with .NET and ML.NET
9. Deployment, DevOps, MLOps, and production

Browse the Data Science for Dummies archive for the parts that remain available.