Introduction to Data Science and Statistics (IDS182) Flashcards

Vocabulary practice flashcards covering fundamental terminology, methodologies, algorithms, descriptive metrics, and inferential statistics from the IDS182 course notes.

Data Science

A multidisciplinary approach that combines mathematics, statistics, specialized analytics, AI/machine learning, and domain expertise to uncover actionable insights from organizational data.

The Data Science Process

A structured methodology for tackling data problems, typically consisting of problem definition, data collection, data preparation, exploratory data analysis, modelling, evaluation, deployment, and monitoring.

Jupyter Notebook

An open-source, browser-based application that provides an interactive workspace combining live code, equations, visualisations, and explanatory Markdown text in a single document.

Reproducibility

The ability for researchers or analysts to replicate study results by executing the same code, data, and analytical workflow.

Bias (in Machine Learning)

The error introduced by oversimplifying a model, causing predictions to deviate systematically from true values and leading to underfitting.

Variance (in Machine Learning)

The error introduced by a model's over-sensitivity to small fluctuations in the training dataset, leading to overfitting on unseen data.

Over-fitting

A condition occurring when a machine learning model is overly complex and learns noise alongside true patterns, performing exceptionally well on training data but poorly on new data.

Under-fitting

A condition occurring when a model is too simple to capture underlying patterns, resulting in poor performance on both training and test datasets.

Supervised Learning

A category of machine learning where an algorithm learns a mapping function from input features ($X$) to known target outputs ($Y$) using labelled data.

Unsupervised Learning

A category of machine learning where an algorithm explores unlabelled data to discover hidden patterns, structures, or groupings without predefined target outcomes.

Classification

A supervised learning problem where the target outcome is categorical or discrete (e.g., pass/fail, churn/no churn).

Regression

A supervised learning task where the target variable is continuous and numerical (e.g., house prices, temperature, revenue).

Clustering

An unsupervised learning technique used to partition unlabelled observations into distinct groups based on feature similarity (e.g., k-Means).

Principal Component Analysis (PCA)

A mathematical technique used in unsupervised learning for dimensionality reduction, transforming correlated features into a smaller set of uncorrelated components.

Precision

A classification accuracy measure calculating the proportion of correctly predicted positive observations relative to all positive predictions made.

Recall

A classification metric assessing the proportion of actual positive cases that the model correctly identified.

F1 Score

A metric combining precision and recall into a single value, particularly useful when evaluating models trained on imbalanced datasets.

Feature Selection

The process of selecting the most relevant variables from a dataset to reduce model complexity, accelerate training, and prevent noise-induced errors.

Regularisation

A technique that penalises complex or oversized model coefficients (e.g., Lasso, Ridge) to reduce overfitting and improve generalisation.

Ensembling

A method that combines outputs from multiple models (e.g., through bagging, boosting, or stacking) to improve predictive accuracy and stability.

Nominal Data

Qualitative categorical data containing distinct groups with no natural or logical order (e.g., department name, gender, blood type).

Ordinal Data

Categorical data that has a clear, meaningful rank order, but lacks uniform, measurable differences between ranks (e.g., job satisfaction ratings).

Interval Data

Quantitative numerical data featuring equal intervals between values but lacking an absolute, true zero point (e.g., temperature in °C).

Ratio Data

Quantitative numerical scale data possessing equal intervals and a meaningful absolute zero, enabling full ratio calculations (e.g., age, income, revenue).

Mean

The arithmetic average of a numerical dataset, calculated by dividing the sum of all observation values by the total sample size ($n$).

Median

The middle value in an ordered dataset, serving as a measure of central tendency resistant to extreme outliers.

Standard Deviation (SD)

A core measure of dispersion representing the average distance or spread of data observations from the sample mean.

Coefficient of Variation (CV)

A relative measure of dispersion calculated as $\text{CV} = \frac{s}{\text{Mean}} \times 100$, useful for comparing variability across datasets with different units.

Confidence Interval (CI)

A computed range of values around a point estimate that is expected to contain the true population parameter at a specified probability level (e.g., 95%).

Null Hypothesis ($H_0$)

A default statistical proposition asserting that no true difference, effect, or relationship exists in the underlying population.

Alternative Hypothesis ($H_1$)

A research claim stating that a statistically meaningful difference, effect, or relationship exists in the population.

p-value

The probability of obtaining a sample result at least as extreme as the observed data, assuming that the null hypothesis ($H_0$) is true.

Type I Error

An error that occurs in hypothesis testing when a researcher incorrectly rejects a true null hypothesis (false positive).

Type II Error

An error that occurs when a researcher fails to reject a false null hypothesis (false negative).

Pearson Correlation Coefficient ($r$)

A standard metric bounded between $-1.0$ and $+1.0$ that measures the direction and linear strength of association between two continuous variables.

Coefficient of Determination ($R^2$)

A statistic ranging from $0$ to $1$ representing the proportion of variance in the dependent target variable explained by a regression model.

t-test

An inferential statistical test used to evaluate whether the means of two distinct groups differ significantly from each other.

ANOVA (Analysis of Variance)

An inferential statistical procedure used to compare sample means across three or more independent groups to test for significant variance.

Model Drift

The progressive decline in a machine learning model's predictive accuracy caused by changing real-world data patterns over time.

Algorithmic Fairness

An ethical principle ensuring that predictive analytics models and algorithms are systematically audited to prevent bias or discrimination against protected groups.