Data Understanding & Preparation

Dataset Playground

Start where every data-science workflow should start: understand the rows, columns, variable types, missingness and target before fitting anything.

1 · Dataset 2 · Preprocessing 3 · Feature engineering 4 · Dimensionality reduction 5 · Class imbalance 6 · Data leakage

Lab concept guide

What to observe while you experiment

Before modelling, the dataset must be understood as a measurement table: what one row represents, what each column means, which values are missing, which fields are targets/identifiers, and whether the table contains duplicates or impossible values.

MechanismInspect row grain, schema, dtypes, missingness, distributions and target balance before choosing transformations or models.

Failure modeTreating a CSV/DataFrame as model-ready merely because it loads successfully.

VerificationCross-check row/column counts, dtypes and missing-value summaries against several raw records and the stated data dictionary.

Experiment deliberately

Choose one dataset and write a five-line data contract: row unit, target, key columns, variable types and one quality risk. Then verify each item in the lab.

Built-in dataset Target Test size 20% Random seed Stratify when possiblePreserve categorical target proportions.

CSV stays in your browser. Uploads are parsed locally and are not sent anywhere.

Loading built-in dataset…

Target distribution

Check class balance or target shape before modelling.

Feature selection

Select which columns would be available to a model. The target is automatically excluded.

Dataset Playground

What to observe while you experiment

Target distribution

Feature selection

Column profile

Row preview