EDA

Exploratory Data Analysis. Examining data structure, quality, and patterns before modelling.

Correlation

A statistical measure of how two variables move together. Does not imply causation.

Causation

One variable producing changes in another. Established through experiments or causal inference methods.

p-value

The probability of observing results this extreme if the null hypothesis were true. Not the probability the alternative is true.

Overfitting

When a model memorises training data and generalises poorly to new data.

Underfitting

When a model is too simple to capture the real patterns in the data.

Bias-variance tradeoff

The tension between a model being too simple (high bias) and too sensitive to training data (high variance).

Feature engineering

Creating or transforming input variables to make a model more accurate or interpretable.

Bayes theorem

P(H|E) = P(E|H) * P(H) / P(E). The rule for updating beliefs given evidence.

A/B test

A randomised experiment assigning users to control and treatment conditions to measure the causal effect of a change.

Regularisation

A technique for reducing overfitting by penalising model complexity during training.

DPDPA

Digital Personal Data Protection Act 2023. India's framework governing how personal data is collected, used, and processed.