Data Science and AI deck

Mean

Average of values sum divided by count.

Median

Middle value when data is sorted.

Mode

Most frequent value in a dataset.

Range

Max minus min in a dataset.

Variance

Average squared distance from the mean.

Standard deviation

Square root of variance; typical spread from the mean.

Correlation

Strength and direction of a linear relationship.

Correlation vs causation

Correlation does not prove one variable causes another.

Conditional probability P(A|B)

Probability of A given B occurred.

Independent events

A and B are independent if P(A|B)=P(A).

Normal distribution

Bell-shaped distribution defined by mean and standard deviation.

Expected value

Weighted average outcome based on probabilities.

Discrete variable

Counts or separate values like number of clicks.

Continuous variable

Any value in a range like time or temperature.

Histogram use

Shows distribution of a numeric variable.

Scatterplot use

Shows relationship between two numeric variables.

Boxplot use

Shows median quartiles spread and potential outliers.

Data cleaning purpose

Fix errors and improve data quality for analysis and models.

Common data issues

Missing values duplicates inconsistent formats outliers bad sources.

Structured data

Fixed schema like rows and columns in tables.

Unstructured data

Text images audio video without a fixed schema.

Semi-structured data

JSON or XML with flexible fields.

Primary key

Uniquely identifies a row in a table.

Foreign key

References a primary key in another table to link tables.

SQL SELECT

Returns columns and rows from a table.

SQL WHERE

Filters rows by condition.

SQL GROUP BY

Groups rows for aggregation like COUNT SUM AVG.

SQL HAVING

Filters grouped results after aggregation.

INNER JOIN

Returns rows with matching keys in both tables.

Pandas used for

Data wrangling cleaning and analysis with DataFrames.

NumPy used for

Fast numerical arrays and mathematical operations.

PyTorch used for

Building and training deep learning models.

Generative AI

AI that produces new content like text images code.

LLM limitation

Can hallucinate and be confidently wrong.

Train validation test split

Train learns; validation tunes; test evaluates generalization.

Supervised learning

Learns from labeled examples to predict targets.

Unsupervised learning

Finds structure in unlabeled data like clusters.

Reinforcement learning

Learns actions through rewards and penalties.

Overfitting

Great training performance but poor new-data performance.

Underfitting

Model too simple; poor performance even on training data.

Confusion matrix

Counts TP FP TN FN for classification.

Precision

TP divided by TP plus FP.

Recall

TP divided by TP plus FN.

Linear regression

Predicts a continuous numeric value.

Logistic regression

Predicts probability for classification.

Decision tree

Splits data by feature rules to classify or predict.

K-means clustering

Groups data into k clusters by similarity.

Algorithmic bias

Systematic unfair outcomes due to data or model choices.

Ethical AI practice

Inclusive data, bias audits, transparency, accountability.