Import and prepare a dataset

Choose a data source

Start by adding a tabular dataset to the current project.

Ounias uses datasets as the starting point for analysis. The imported dataset remains associated with the project in which it was created.

• Local CSV or XLSX file
• Google Drive
• Google Sheets

Review the imported dataset

Confirm that the expected rows and columns have been loaded before defining their analytical roles.

The dataset preview provides an initial check of the imported structure before Ounias begins configuring it for analysis.

Define feature roles

Review the automatically detected column types and specify how each column should participate in analysis.

Ounias attempts to identify likely feature types automatically. Review these assignments before continuing and exclude columns that should not participate in analysis.

• Continuous
• Categorical
• Binary
• Identifier
• Supervised target
• Longitudinal entity
• Observation index
• Observation datetime

Validate the configuration

Ounias checks whether the selected feature definitions are compatible with the underlying data.

Validation surfaces inconsistencies between the configured role of a column and the values actually present in it.

Resolve detected issues before continuing so downstream analytical pipelines receive a consistent dataset.

Prepare the analytical data

Ounias applies the preprocessing required to make configured features usable by downstream methods.

• Missing-value handling
• Cleaning and normalization where required
• Conversion of supported feature types into analytical representations
• Method-compatible transformations

Profile the dataset

Dataset-level diagnostics are calculated once ingestion is complete.

Profiling describes structural characteristics of the dataset and evaluates compatibility with supported clustering, supervised, and embedding methods.

Relevant compatibility information is reused later when Ounias presents methods and configuration choices in the Run Launcher.

• Dataset structure and missingness
• Correlation and outlier characteristics
• Clustering suitability
• Embedding characteristics
• Supervised-analysis compatibility
• Longitudinal readiness where applicable

Start analyzing

Once ingestion is complete, the dataset becomes available to Ounias analytical workflows.

You can now create individual runs through the Run Launcher or use guided evaluation to explore several analytical approaches.

Generated runs, metrics, artifacts, explanations, and later variations remain associated with this dataset.