Data Science Vocabulary in English

A dataset is a structured collection of data — typically a table of rows and columns — used specifically for analysis, training, or research. It is often a static file in CSV, JSON, or Parquet format. A database is a dynamic, organised system for storing, querying, and managing large volumes of data, typically accessed via SQL. In data science, data is often extracted from databases to create datasets for analysis or model training.

Overfitting occurs when a machine learning model learns the training data too precisely — including its noise and random fluctuations — so it performs well on training data but poorly on new, unseen data. The model has memorised rather than generalised. Underfitting is the opposite: the model is too simple to capture the patterns in the data, performing poorly even on the training data. Good model development aims for the middle ground: a model that generalises well to new data without either memorising or oversimplifying.

Correlation means that two variables tend to change together — when one goes up, the other goes up (positive correlation) or down (negative correlation). Causation means that one variable directly causes the change in the other. A famous example: ice cream sales and drowning rates are correlated (both rise in summer), but ice cream does not cause drowning — both are caused by hot weather. Data scientists must be careful not to confuse correlation with causation when interpreting results.

A neural network is a computational model loosely inspired by the structure of the human brain. It consists of layers of interconnected nodes (neurons) that process input data, adjust the strength of connections (weights) based on errors, and learn to recognise patterns over many iterations. During training, the network makes predictions, compares them to the correct answers, calculates the error, and adjusts the weights slightly to reduce the error. This process (backpropagation) is repeated across thousands or millions of examples.

A data pipeline is an automated sequence of steps that moves and transforms data from one place to another. A typical pipeline might: extract raw data from a database or API, clean and validate the data, transform it (normalise, encode, engineer features), and load it into a data warehouse or pass it to a machine learning model for training or inference. Pipelines automate repetitive data preparation tasks and ensure consistency and reproducibility in data workflows.

Training data is the dataset used to fit a machine learning model — the model learns patterns from this data by adjusting its parameters. Validation data (or the validation set) is a separate portion of the data held back from training, used to evaluate the model's performance during development. This helps detect overfitting and guides choices about model architecture and hyperparameters. A third set — the test set — is used only at the very end to give a final, unbiased estimate of model performance.

Normalisation (also called feature scaling) is the process of transforming data values to a common scale without distorting differences in the ranges. For example, if one feature is “age” (ranging from 18 to 90) and another is “salary” (ranging from 20,000 to 200,000), the salary values would dominate calculations simply because they are larger numbers. Normalising both features to a range of 0 to 1 ensures that each feature contributes appropriately to the model.

Bias in machine learning refers to systematic errors in a model that cause it to make consistently inaccurate or unfair predictions. Model bias often reflects biases in the training data — for example, a hiring algorithm trained on historical data may discriminate against women if historical hiring was male-dominated. Bias is a serious ethical and technical problem in AI, affecting areas like facial recognition, credit scoring, criminal justice prediction, and medical diagnosis.

Supervised learning involves training a model on labelled data — each example has an input and a known correct output. The model learns to map inputs to outputs, enabling prediction on new data. Regression and classification are both supervised learning tasks. Unsupervised learning involves finding patterns in unlabelled data — there are no correct answers given to the model. Clustering is a common unsupervised technique, grouping similar data points without predefined categories.

Data science job descriptions are full of technical vocabulary: algorithm, model, pipeline, training, validation, regression, clustering, feature engineering, and inference appear regularly. In technical interviews, candidates are expected to explain these concepts clearly and correctly in English. Understanding the vocabulary also helps non-data-scientist professionals — managers, marketers, analysts — collaborate effectively with data teams, evaluate reports critically, and participate in evidence-based decision-making.