Flashcards ISBA 1 Final Exam Review: Unsupervised Learning, Text Mining, Prescriptive Analytics & Ethics

Unsupervised Learning

Uses unlabeled data to discover hidden patterns and structures.

Supervised Learning

Uses labeled data to predict outcomes.

Labeled data

Data with a target attribute.

Unlabeled data

Data without a target attribute.

Main goals of Unsupervised Learning

Discover hidden patterns, find clusters, reduce dimensions, detect anomalies, visualize large datasets.

Clustering

Groups similar data points.

Dimensionality Reduction

Reduces the number of features.

Association Rule Mining

Finds relationships between items.

Anomaly Detection

Identifies unusual patterns.

Cluster

A group of similar data points.

Centroid

Center of a cluster, calculated using the mean.

Distance Metric

Measures distance or similarity between data points.

K-means

Groups data based on nearest centroids.

DBSCAN

Density-Based Spatial Clustering of Applications with Noise.

Euclidean Distance

Shortest straight-line distance.

Silhouette Coefficient

Measures cluster cohesion and separation.

Inertia / WCSS

Measures cluster compactness; lower is better.

Curse of Dimensionality

Data becomes sparse as dimensions increase.

PCA

Captures maximum variance.

Support

Frequency of an itemset.

Confidence

Strength of the association between two items.

Lift

Strength of association compared with independence.

Outlier Detection

Finds abnormal data points in training data.

Z-score

Measures standard deviations from the mean; generally, above 3 indicates an outlier.

IQR

Uses Q1 and Q3 to identify extreme values.

Modified Z-score

Uses median and Median Absolute Deviation (MAD).

Isolation Forest

Uses isolation trees.

LOF

Identifies points with significantly lower local density than neighbors.

ABOD

Measures angles between individual points.

Precision

Correctly identified anomalies among flagged instances.

Recall

Actual anomalies successfully detected.

F1 Score

Balances precision and recall.

AUC-ROC

Evaluates true positive rate against false positive rate.

AUC-PR

Evaluates precision and recall.

Text Mining / Text Analytics

Transforms unstructured text into meaningful, structured information for analysis and decision-making.

Automation

Summarization and classification.

Business Value

Understand customer behavior.

Research

Knowledge discovery.

Large Volume

Processes large amounts of text.

Machine Learning

Learns from data and experience.

Data Mining

Discovers patterns in datasets.

NLP

Enables computers to understand human language.

Information Retrieval

Finds and ranks relevant information.

Computational Linguistics

Uses algorithms to analyze language.

Statistics

Analyzes and interprets data.

Tokenization

Splits text into words or phrases.

Lowercasing

Converts text to lowercase.

Stopword Removal

Removes common words like 'the' and 'is'.

Stemming

Reduces words to root form; may produce nonwords.

Lemmatization

Converts words into dictionary forms using grammar.

Noise Removal

Removes punctuation, numbers, HTML tags, and emojis.

PoS Tagging

Identifies parts of speech.

Word Sense Disambiguation

Determines the correct meaning of a word based on context.

Full Parsing

Produces a parse tree.

Chunking

Identifies partial grammatical structures.

Bag of Words (BoW)

Represents text using word frequencies.

TF-IDF

Highlights important words and reduces the weight of common terms.

Word Embeddings

Captures semantic meaning using vectors.

Contextual Embeddings

Captures word meaning based on surrounding words.

NER

Identifies people, organizations, locations, dates, and times.

Relation Extraction

Identifies relationships between entities.

Event Extraction

Identifies events in text.

Text Classification

Assigns categories to documents.

Sentiment Analysis

Detects positive, negative, or neutral emotions.

Topic Modeling

Discovers hidden themes.

Text Summarization

Produces summaries.

Text Generation

Creates new text.

Classification Evaluation

Accuracy, Precision, Recall, F1-score.

Topic Modeling Evaluation

Coherence Score, Human Evaluation.

Clustering Evaluation

Silhouette Score.

Language Models Evaluation

Perplexity.

Summarization / Translation Evaluation

BLEU / ROUGE.

Prescriptive Analytics

Uses data, mathematical models, and algorithms to recommend the best possible action or decision.

Descriptive Analytics

What happened?

Diagnostic Analytics

Why did it happen?

Predictive Analytics

What will happen?

Optimization Techniques

Finds the best possible solution given practical constraints.

Linear Programming (LP)

Optimizes a linear objective with linear constraints; uses continuous decision variables.

Integer Programming

Uses whole-number decisions.

Mixed-Integer Programming

Combines integer and continuous variables.

Nonlinear Programming

Uses nonlinear objective functions or constraints.

Simulation Techniques

Creates digital models of real-world systems.

Monte Carlo

Uses random sampling and probability.

Discrete-Event Simulation

State changes at specific events.

Continuous Simulation

State variables change continuously.

Agent-Based Simulation

Studies actions and interactions of individual agents.

Decision Analysis

Systematic approach to making decisions under uncertainty.

Decision Trees

Visual branching diagrams showing decisions and possible outcomes.

Cost-Benefit Analysis

Compares expected costs and benefits.

Scenario Planning

Prepares strategies for possible future scenarios.

Utility Theory

Assigns values based on preference or satisfaction.

Heuristic

Problem-specific shortcuts or rules of thumb; does not guarantee optimality.

Meta-Heuristic

General strategies that guide and improve simpler heuristics.

Reinforcement Learning

A machine learning technique where an agent learns optimal actions through trial and error.

Dynamic pricing

Application of reinforcement learning.

Inventory replenishment

Application of reinforcement learning.

Robotics

Application of reinforcement learning.

Supply chain routing

Application of reinforcement learning.

Prescriptive Analytics Process

A series of steps to define problems, collect data, build models, and implement recommendations.

Define the Problem

Identify objectives, constraints, and stakeholders.