Business Intelligence, Analytics, Data Science, and AI Midterm Review

Comprehensive vocabulary flashcards covering the foundations of business intelligence, the three analytics tiers, statistical modeling, big data technologies, data warehousing architectures, AI/robotics, and data mining algorithms.

Business Intelligence (BI)

A conceptual framework for managerial decision support that combines architecture, databases (or data warehouses), analytical tools, applications, and methodologies.

Decision Support Systems (DSS)

Computerized systems that utilize data, models, and sometimes knowledge management to find solutions for semistructured and some unstructured problems.

Descriptive Analytics

An earlier phase in the analytics continuum that deals with describing data to answer the questions of what happened and what is happening using reporting, dashboards, scorecards, and data warehousing.

Predictive Analytics

A business analytical approach toward forecasting demand, problems, or opportunities using statistical analysis, data mining, text mining, and web/media mining to answer what will happen and why.

Prescriptive Analytics

A branch of business analytics (also known as decision or normative analytics) that uses optimization, simulation, and decision modeling to determine the best possible decisions and actions.

Online Transaction Processing (OLTP)

A transaction processing system primarily responsible for capturing and storing operational data related to day-to-day business functions like ERP, CRM, and SCM.

Online Analytical Processing (OLAP)

An information system architecture that enables end users to query data warehouses, conduct multidimensional analysis (such as slice-and-dice or drill-down), and generate results in seconds.

Analytics Ecosystem

A classification of sectors, technology/solution providers, and industry participants organized into 11 clusters surrounding the analytics user organization.

Gordon Moore

Co-founder of Intel whose observation (Moore's Law) tracks the exponential growth of processing power, with transistor counts rising from $2.3 \times 10^3$ in $1971$ to projected levels exceeding $134 \times 10^9$ by $2026$.

Big Data

Data that cannot be stored or processed easily using traditional tools due to its scale and characteristics, specifically defined by volume, variety, velocity, veracity, and value.

Artificial Intelligence (AI)

The ability of machines or computer systems to perform tasks typically requiring human intelligence, such as visual perception, pattern recognition, and autonomous decision-making.

Huggable

A social robot equipped with AI to understand facial expressions, temperament, gestures, and touch to provide social and emotional support to children in hospital treatments.

Robot

A physical machine or device that operates with AI cooperation to accomplish responsibilities, shifting from automation of repetitive tasks to self-initiated autonomy.

AI Chatbot

A computerized conversational service between humans and machine interfaces, powered by natural language processing (NLP), natural language generation, and machine learning over cloud services.

Data Preprocessing

A multi-step process that converts raw data into an analytics-ready state through four major phases: data consolidation, data cleaning, data transformation, and data reduction.

Structured Data

Data formatted into organized tables with rows and columns, making it directly interpretable and processable by computer software.

Unstructured Data

Data that does not have a predetermined format, typically found in textual documents, audio files, images, or video streams.

Hadoop

An open-source software framework created by Doug Cutting at Yahoo! designed to store and analyze massive volumes of distributed, unstructured data across inexpensive commodity hardware.

MapReduce

A distributed computing technique in Hadoop that processes large data sets by first applying a Map function to parse and categorize elements, followed by a Reduce function to aggregate results.

Arithmetic Mean

A central tendency measure calculated as the simple average of all numerical sample values: $\bar{x} = \frac{\sum_{i=1}^n x_i}{n}$.

Median

A central tendency metric in statistics identifying the value situated at the 50th percentile (the middle number) of a sorted data set.

Mode

A measure of central tendency representing the most frequently occurring value or observation in a data set.

Variance

A measure of statistical dispersion representing the average squared deviation of data values from their mean: $s^2 = \frac{\sum_{i=1}^n (x_i - \bar{x})^2}{n - 1}$.

Standard Deviation

A dispersion measure assessing the spread of numeric points, calculated as the positive square root of variance: $s = \sqrt{\frac{\sum_{i=1}^n (x_i - \bar{x})^2}{n - 1}}$.

Mean Absolute Deviation (MAD)

A measure of dispersion and accuracy metric calculated as the average absolute deviation of values from the dataset mean.

Box-and-Whiskers Plot

A graphical display showing the central tendency (median, mean) and dispersion (quartiles, min, max, outliers exceeding $1.5$ times the quartile) of a dataset.

Ordinary Least Squares (OLS)

A mathematical optimization method in regression modeling that determines the line of best fit by minimizing the sum of the squared vertical distances between data points and the line.

Logistic Regression

A supervised, probability-based classification algorithm used when the dependent variable is binary or categorical, using the formula $f(y) = \frac{1}{1 + e^{-(\beta_0 + \beta_1 x)}}$.

Data Warehouse (DW)

A physical, centralized repository of integrated, time-variant, nonvolatile, and subject-oriented historical and current relational data organized to support managerial decision-making.

Data Silos

Disconnected, isolated islands of departmental data that have divergent standards, differing qualities, disparate database systems, and conflicting semantics.

Data Mart (DM)

A departmental, small-scale data repository that focuses on a single subject area (such as marketing, finance, or operations) to meet the specific reporting needs of a business unit.

Dependent Data Mart

A specialized subset of an enterprise data warehouse populated directly with data extracted from the central warehouse.

Independent Data Mart

A small, self-contained data mart designed for a strategic business unit or department that loads data directly from operational staging areas without a central data warehouse.

Data Mart Bus Architecture

A data warehouse architecture where individual dimensional data marts are linked together across the enterprise through shared, conformed dimensions.

Hub-and-Spoke Architecture

A corporate information factory architecture where data is first integrated into a centralized, normalized relational warehouse (the hub) before feeding dependent data marts (the spokes).

Centralized Data Warehouse Architecture

An enterprise data warehouse architecture where a single normalized relational warehouse contains all atomic and summarized organizational data, accessed directly by end-user applications without intermediate data marts.

Federated Architecture

A data warehouse design where common data elements are logically or physically integrated on-demand across existing disparate data warehouses, data marts, and legacy systems using metadata mapping.

Enterprise Application Integration (EAI)

A middleware-based integration technology that connects and pushes data among multiple enterprise applications and databases.

Enterprise Information Integration (EII)

An evolving integration technology that provides real-time, on-demand virtual data access and integration from diverse sources including relational databases, web services, and multidimensional cubes.

ETL Process

A data pipeline comprising extraction (reading from disparate source databases), transformation and cleansing (formatting, standardizing, and deduplicating), and loading into a data warehouse or data mart.

Inmon Model

A top-down data warehousing approach pioneered by Bill Inmon that first constructs an enterprise data warehouse in third normal form ($3\text{NF}$) before creating specialized, dependent data marts.

Kimball Model

A bottom-up data warehousing approach created by Ralph Kimball that builds individual dimensional data marts using star schemas and joins them together via conformed dimensions.

Star Schema vs. Snowflake Schema

Star schema features a central fact table surrounded by denormalized, single-table dimensions; snowflake schema extends this design by normalizing dimension tables into multiple related sub-tables.

Data Lake

A large-scale storage repository designed for low-cost ingestion of massive, highly flexible, raw/native unstructured or structured data, typically queried with NoSQL by data scientists.

PERT Chart

A network diagram visualization technique used in project management to illustrate the sequence, relationships, and precedence of dependent tasks from start to completion.

Bullet Graph

A specialized variation of a bar chart that displays a single primary measure alongside a target indicator and background performance ranges (e.g., poor, satisfactory, good).

Dashboard

A visual presentation tool that consolidates KPIs and vital business metrics on a single screen to provide real-time situational awareness and drill-down exploration.

Visual Analytics

An advanced analytical practice combining information visualization (descriptive, past-focused) with predictive analytics (forecasting, future-focused) to enhance decision-making.

Balanced Scorecard (BSC)

A strategic performance management methodology that evaluates organization objectives across four balanced perspectives: financial, customer, internal business process, and learning/growth.

Six Sigma

A disciplined, data-driven process improvement methodology aimed at reducing defects to near zero (fewer than $3.4$ defects per million opportunities) using the DMAIC cycle.

Data Mining

A computational process at the intersection of statistics, AI, and machine learning that extracts hidden patterns, novel correlations, and actionable knowledge from large databases.

CRISP-DM

The Cross-Industry Standard Process for Data Mining, a six-step methodology starting with business understanding and data understanding (which consume over $80\%$ of project time) and concluding with deployment.

SEMMA

A standardized data mining process developed by the SAS Institute whose acronym stands for Sample, Explore, Modify, Model, and Assess.

Classification

A supervised data mining technique that analyzes historical records containing known labels to construct a model capable of assigning future categorical outcomes.

Clustering

An unsupervised data mining task that partitions unlabeled records into natural, mutually homogeneous segments based on calculated similarity or distance measures.

Gini Index

A purity metric used to evaluate branch splits in decision trees; if all records in a partition belong strictly to one class, the Gini index value is $0$.

Information Gain

The attribute splitting criterion utilized in the ID3 decision tree algorithm, derived by measuring the reduction in entropy (uncertainty).

Association Rule Mining

A pattern discovery technique (popularly known as market-basket analysis) that identifies items, events, or behaviors that frequently co-occur within transactional records.

Support

In association analysis, the proportion of total transactions in a dataset that contain all items specified in both the antecedent and consequent of a rule.

Confidence

In association analysis, the conditional probability that the right-hand side (consequent) appears in a transaction given that the left-hand side (antecedent) is already present.

k-Fold Cross-Validation

A predictive model assessment technique where data is randomly split into $k$ equal subsets; the model is iteratively trained on $k - 1$ subsets and tested on the remaining fold $k$ times to compute average accuracy.

Bootstrapping

A validation sampling technique where a fixed number of instances are drawn with replacement from the dataset for training, while unsampled records serve as the test set.