good one

Anscombe’s quartet

A set of four datasets that have nearly identical statistical summaries (mean, varianData science process

Information

Processed or organized data that has meaning and context.

Knowledge

Information interpreted and applied to make decisions or take action.

Data Science

An interdisciplinary field that uses statistics, mathematics, programming, and domain knowlKnowledge Engineering

Median

Middle value of a sorted dataset.

Mode

Most frequently occurring value.

Range

Difference between the largest and smallest values.

Variance

Average squared deviation from the mean.

Standard deviation

Square root of variance; represents spread in original units.

Interquartile range (IQR)

Difference between Q3 and Q1; robust measure of spread.

Skewness

Measure of asymmetry in a distribution.

Kurtosis

Measure of tail heaviness compared to a normal distribution.

Z-score

Standardized value calculated as (x − µ)/σ.

Expected value

Long-term average outcome of a random variable.

Population

Entire group of interest in a study.

Sample

Subset drawn from a population to estimate characteristics.

Parameter

Numerical summary of a population (µ, σ).

Statistic

Numerical summary of a sample (x

Law of large numbers

Sample mean approaches population mean as sample size increases.

Central limit theorem

Distribution of sample means approaches normal as sample size increases.

Correlation

Measures how strongly two variables move together.

Correlation vs causation

Correlation indicates association; causation requires proof that one variable cPearson correlation

Spearman correlation

Rank-based correlation robust to outliers and nonlinear relationships.

Covariance

Measures whether two variables increase or decrease together.

Outlier

Data point far from the rest of the dataset.

Probability

Likelihood of an event occurring.

Sample space

All possible outcomes of an experiment.

Conditional probability

Probability of event A occurring given event B.

Joint probability

Probability of two events occurring together.

Bayes’ theorem

Formula for updating probabilities using evidence.

Prior probability

Belief about an event before observing data.

Posterior probability

Updated belief after observing evidence.

Likelihood

Probability of observing data given parameters.

Normal distribution

Bell-shaped distribution defined by mean and standard deviation.

Empirical rule

68%, 95%, and 99.7% of data fall within 1, 2, and 3 standard deviations.

Binomial distribution

Distribution for number of successes in fixed trials.

Poisson distribution

Distribution for counts of events occurring in a time interval.

Exponential distribution

Distribution describing time between events.

Bernoulli distribution

Distribution for a single success/failure trial.

Structured data

Data organized into rows and columns.

Unstructured data

Data without predefined structure (text, images, audio).

Numeric data

Data represented with numbers.

Categorical data

Data represented as categories or labels.

Discrete variable

Variable with countable values.

Continuous variable

Variable with infinite values within a range.

Random variable

Variable whose value depends on random outcomes.

Bar graph

Used to compare categories.

Histogram

Shows distribution of numerical data.

Line graph

Shows trends over time.

Scatter plot

Shows relationship between two variables.

Box plot

Displays quartiles, spread, and outliers.

Pie chart

Shows proportions of a whole.

Heatmap

Uses color to represent data intensity or correlation.

Data visualization

Graphical representation of data to reveal patterns.

Data cleaning

Fixing errors, removing duplicates, handling missing values.

Data wrangling

Organizing raw data for analysis.

Data transformation

Converting data into a useful format.

Data quality issues

Problems such as missing values, duplicates, or noise.

Missing data mechanisms

MCAR, MAR, MNAR.

Imputation methods

Methods to fill missing values (mean, median, KNN, regression).

Feature engineering

Creating new features from existing data.

Feature scaling

Adjusting data ranges using normalization or standardization.

Feature encoding

Converting categorical variables to numerical format.

Feature selection

Choosing relevant variables for modeling.

Dimensionality reduction

Reducing number of variables while preserving information.

Principal component analysis (PCA)

Linear dimensionality reduction technique.

t-SNE / UMAP

Nonlinear dimensionality reduction for visualization.

Autoencoders

Neural networks used for dimensionality reduction.

Data leakage

When training data accidentally includes information from test data.

Machine learning

Algorithms that learn patterns from data.

Supervised learning

Learning using labeled data.

Unsupervised learning

Finding patterns in unlabeled data.

Reinforcement learning

Learning by rewards and penalties.

Training dataset

Data used to train a model.

Validation dataset

Data used to tune model parameters.

Test dataset

Data used to evaluate final model performance.

Bias-variance tradeoff

Balance between underfitting and overfitting.

Overfitting

Model memorizes training data but performs poorly on new data.

Underfitting

Model too simple to capture patterns.

Regularization

Techniques that prevent overfitting.

Early stopping

Stopping training when validation performance stops improving.

Ensemble methods

Combine multiple models for better performance.

Bagging

Training models independently on bootstrapped datasets.

Boosting

Sequentially improving models by focusing on errors.

Stacking

Combining predictions of multiple models.

Linear regression

Predicts continuous values using a best-fit line.

Multiple linear regression

Regression using multiple predictors.

Logistic regression

Classification model predicting probabilities.

K-nearest neighbors (KNN)

Classifies based on nearby training examples.

Naive Bayes

Probabilistic classifier assuming feature independence.

Decision tree

Tree structure splitting data based on conditions.

Random forest

Ensemble of decision trees using bagging.

Gradient boosting

Sequential ensemble method focusing on residual errors.

Support vector machine (SVM)

Classifier maximizing margin between classes.

K-means clustering

Groups data into k clusters.

Hierarchical clustering

Creates nested clusters represented as dendrograms.

DBSCAN

Density-based clustering algorithm.

Topic modeling (LDA)

Extracts topics from text documents.

Accuracy

Proportion of correct predictions.

Precision

True positives ÷ predicted positives.