ITEC 4230 NOTES

Data Mining (KDD)

Extraction of interesting patterns or knowledge from a huge amount of data.

Classification

Find a model for the class as a function of other attributes to assign a class to unseen records.

Clustering

Grouping similar data objects together without predefined class labels.

Association Rule Discovery

Producing dependency rules predicting the occurrence of an item based on other items' occurrences.

Sequential Pattern Discovery

Finding rules that predict strong sequential dependencies among different events.

Regression

Predicting a value of a continuous variable based on other variables.

Deviation/Anomaly Detection

Detecting significant deviations from normal behavior.

Descriptive data mining

Mining that characterizes or summarizes existing patterns in the data.

Predictive data mining

Mining that forecasts unknown or future values.

Data object

A representation of an entity, also called a sample, example, or data point.

Attribute

A data field representing a characteristic of a data object.

Nominal attribute

Categories or 'names of things' with no meaningful order.

Binary attribute

A nominal attribute with only two states (0 and 1).

Ordinal attribute

Values have a meaningful order/ranking, but magnitude between values is unknown.

Interval-scaled attribute

Measured on a scale of equal-sized units with order, but no true zero-point.

Ratio-scaled attribute

Has an inherent zero-point, allowing meaningful ratios between values.

Discrete attribute

Has a finite or countably infinite set of values.

Continuous attribute

Has real numbers as its values.

Mean

The algebraic average of a set of values.

Median

The middle value or average of the two middle values in a dataset.

Mode

The value that occurs most frequently in the data.

Quartiles (Q1, Q3)

The 25th and 75th percentiles of the data.

Inter-Quartile Range (IQR)

The range between the 25th percentile (Q1) and the 75th percentile (Q3).

Outlier

A value significantly higher or lower than the IQR from the quartiles.

Variance

A measure of dispersion computed as the average squared deviation from the mean.

Standard deviation

The square root of the variance.

Boxplot

A graphic display showing quartiles and outliers of data.

Histogram

A graphical display of frequencies in bar form.

Quantile plot

A plot pairing each value with its corresponding percentile.

Quantile-quantile (Q-Q) plot

Graphs quantiles of one distribution against another.

Scatter plot

A plot where each pair of values is treated as a coordinate pair.

Similarity

A numerical measure of how alike two data objects are.

Dissimilarity (distance)

A measure of how different two data objects are.

Data matrix

A structure storing data points with dimensions.

Dissimilarity matrix

A matrix registering distances between data points.

Minkowski distance

A parameterized distance measure between two p-dimensional objects.

Metric

A distance measure satisfying positive definiteness, symmetry, and triangle inequality.

Cosine similarity

A measure based on the cosine of the angle between two vectors.

Structured data

Data fitting into a fixed row/column format.

Unstructured data

Data without a fixed shape or format.

Business process

A sequence of activities leading to an outcome of value.

Information system

Components managing information for decision-making and analysis.

Data

Raw facts in their original form.

Information

Data shaped into a meaningful, useful form.

Entity

A person, place, or thing about which information is stored.

Field

A column in a database representing specific values.

Record

A group of related fields in a database.

File

A group of records of the same type.

Database

A group of related files.

Data model

Concepts used to describe the structure of a database.

Conceptual data model

A high-level model close to how users perceive data.

Implementation data model

A model used by commercial DBMS, between conceptual and physical.

Physical data model

A model describing how data is stored in a computer.

Database schema

Description of a database's structure, data types, and constraints.

Data-Driven Decision Making (DDDM)

A process involving data collection and analysis for business strategy.

Machine Learning

An AI method enabling systems to learn from data.

Machine-learning model

Output from a learning algorithm trained with data.

Linear regression

Estimating one variable from another when they are linearly dependent.

Logistic regression

A regression model with a categorical dependent variable.

Data Analytics

The science of analyzing raw data for conclusions.

Descriptive analytics

Analytics answering 'What happened?' with historical data.

Diagnostic analytics

Analytics identifying the root cause of a problem.

Predictive analytics

Statistical methods predicting future behaviors based on trends.

Prescriptive analytics

Methods generating recommendations based on predictive models.

Data Warehouse

Collects and manages data from varied sources for business insights.

Data Mart

A simplified data warehouse focused on specific subjects.

Data cleaning

Filling missing values and resolving inconsistencies in data.

Data integration

Combining data from multiple sources into a coherent store.

Data reduction

Obtaining a reduced representation of a dataset.

Data transformation

Mapping an entire set of attribute values to a new set of values.

Noise

Random error or variance in measured data.

Schema integration

Integrating metadata from different sources.

Entity identification problem

Identifying records referring to the same entity from different sources.

ETL

Extraction, Transformation, and Loading tools for data transformations.

Data scrubbing

Detecting and correcting errors using domain knowledge.

Data auditing

Analyzing data to discover rules that reveal outliers.

Chi-square (χ²) test

A correlation measure for nominal data.

Pearson's correlation coefficient (r)

A measure of linear correlation between two numeric variables.

Covariance

How two variables change together.

Curse of dimensionality

Sparsity and ineffectiveness of distance measures as dimensionality increases.

Principal Component Analysis (PCA)

Projection method capturing the largest variation in the data.

Redundant attribute

An attribute duplicating information in another attribute.

Irrelevant attribute

An attribute that is non-contributory to the data mining task.

Wavelet transform

Decomposes a signal to preserve distance between objects at different resolutions.

Parametric data reduction methods

Methods estimating a model's parameters for storage.

Non-parametric data reduction methods

Methods not assuming a model.

Data compression

Encoding data to a smaller size, either lossless or lossy.

Normalization

Scaling attribute values to a specified range.

Discretization

Dividing a continuous attribute's range into intervals.

Concept hierarchy

Hierarchy of attribute values for multi-level data viewing.

Data Warehouse (Inmon's definition)

A subject-oriented, integrated collection of data supporting decision-making.

Subject-oriented

Organized around major subjects rather than daily operations.

Integrated

Constructed from integrating multiple data sources.

Time-variant

Covers a longer time frame than operational systems.

Nonvolatile

Only initial loading and reading occurs in a separate store.

Data warehousing

The process of constructing and utilizing data warehouses.

OLTP (On-Line Transaction Processing)

System oriented towards daily operational transactions.

OLAP (On-Line Analytical Processing)

System oriented towards decision support and multidimensional queries.

Enterprise warehouse

Collects all organizational information on major subjects.

Data mart

A subset of data valuable to a specific group.