Correlation Analysis

This method computes pairwise correlations between numeric features using Pearson and/or Spearman and/or Kendall correlation. The output is a correlation matrix that helps you spot relationships, redundancy, and potential multicollinearity in your dataset.

How to run (UI)

  1. Open the link for the correlation along with dataset-id parameter (e.g., https://data-valuation-ui.aidapt.s5projects.eu/correlation?dataset-id=XXX)

  2. Choose the correlation method (Pearson and/or Spearman and/or Kendall)

    Specifically

    • Pearson for linear relationships

    • Spearman for monotonic relationships and better robustness to non linear patterns and outliers

    • Kendall for Kendall relationships

    • Select any combination if you want to compare the two views

  3. Click Run Correlation Analysis

    UI-correlation
    Figure 1. Correlation Analysis UI output.

  4. Review the correlation matrices

    • Heatmap colors indicate the strength of the relationship

    • Values on the diagonal are always 1 because each feature is perfectly correlated with itself

  5. Optionally download the results as CSV for further analysis

Interpretation

A correlation value ranges from -1 to +1

  • Values close to +1 indicate a strong positive relationship

  • As one feature increases, the other tends to increase

  • Values close to -1 indicate a strong negative relationship

  • As one feature increases, the other tends to decrease

  • Values close to 0 indicate a weak or no relationship

Notes and good practice

  • High absolute correlation, for example |r| > 0.8, can indicate redundant features and possible multicollinearity for linear models

  • Correlation does not imply causation. Use it as an exploratory signal rather than proof of a causal effect

  • Non numeric and constant features are ignored in this analysis