Unit 2: Introduction to data mining
Data Mining for Business Decisions notes · PTU syllabus (MBA 941-18)
On this page
Unit summary
Data mining discovers useful patterns hidden in large data sets. This unit covers the data mining process and functionalities, text, web, pattern, sequence and context-based mining, data visualisation, predictive vs descriptive data mining, and the need for data analytics in business intelligence.
After this unit you can
- Explain the data mining process (KDD and CRISP-DM)
- Explain data mining functionalities
- Explain text, web, sequence and pattern mining
- Distinguish predictive and descriptive mining and explain the role of analytics in BI
PTU syllabus topics
- Data mining process and functionalities
- text/web/pattern/sequence/context-based mining
- data visualization
- predictive vs descriptive data mining
- need for data analytics in business intelligence
Topic 1
Meaning of data mining
Data mining is a step in knowledge discovery in databases (KDD): data cleaning, integration, selection, transformation, mining, pattern evaluation and knowledge presentation. Main tasks are classification, prediction, clustering, association and outlier detection. Data mining is the process of discovering interesting, non-obvious and useful patterns, relationships and trends in large data sets using statistics, machine learning and database techniques.
- 1
Data cleaning
- 2
Data integration
- 3
Data selection
- 4
Data transformation
- 5
Data mining
Apply algorithms
- 6
Pattern evaluation
- 7
Knowledge presentation
Example
Market basket analysis: a supermarket mines its billing data and finds that customers who buy bread often buy butter and eggs. It places them nearby and creates combo offers.
Exam tip
Distinguish data mining (discovering patterns) from querying (retrieving known facts).
Topic 2
The CRISP-DM process
CRISP-DM (Cross-Industry Standard Process for Data Mining) has six phases: business understanding, data understanding, data preparation, modelling, evaluation and deployment. The process is iterative — teams often move back and forth between phases.
- 1. Business understanding: Define objectives and success criteria
- 2. Data understanding: Collect, describe, explore data
- 3. Data preparation: Select, clean, construct, integrate, format
- 4. Modelling: Choose and run techniques
- 5. Evaluation: Check results against business goals
- 6. Deployment: Put the model to use and monitor
- About 60–80% of a project's effort typically goes into data understanding and preparation.
Example
CRISP-DM for churn: a telecom firm wants to reduce customer churn (business understanding), explores usage and complaint data (data understanding), cleans it (data preparation), builds a decision tree (modelling), checks its accuracy on new data (evaluation), and uses it to send retention offers (deployment).
Exam tip
Data preparation usually takes the most time — often more than half the project.
Topic 3
Data mining functionalities
Data mining can do several kinds of tasks. Each answers a different business question, from summarising data to predicting outcomes and finding unusual cases.
Characterisation and discrimination
Summarise a class, compare classes
Association
Items that occur together
Classification
Assign cases to predefined classes
Prediction (regression)
Estimate a numeric value
Clustering
Find natural groups
Outlier (anomaly) detection
Spot unusual cases — fraud
Evolution and trend analysis
Patterns over time
Example
In a supermarket: association finds that bread and butter are bought together. Classification predicts whether a customer will respond to an offer. Clustering groups customers into segments with no labels given. Outlier analysis flags an unusual card transaction.
Topic 4
Text, web, pattern, sequence and context-based mining
- Text mining: extracting information from unstructured text — sentiment analysis of reviews, topic modelling, classification of complaints, keyword extraction.
- Web mining: content mining (page contents), structure mining (links between pages — PageRank) and usage mining (clickstreams, logs).
- Pattern mining: frequent itemsets and association rules.
- Sequence mining: ordered patterns over time — customers who buy a phone then buy a cover within a week.
- Context-based mining: patterns that depend on context — time, location, device, weather — used in recommendations and location-based offers.
Example
A telecom firm mines complaint text and finds "network drop" in one district spiking after a tower upgrade — a pattern no structured report showed.
Topic 5
Data visualisation in data mining
- Purpose: explore data before modelling, communicate patterns, evaluate models.
- Tools: scatter plot matrices, heat maps, parallel coordinates, tree diagrams, cluster plots, lift and ROC charts, network graphs.
Example
Spotting a pattern visually: a scatter plot of customer age against spending reveals two clear groups — young, low-spending students and middle-aged high spenders — suggesting clustering before any model is built.
Exam tip
Choose the chart for the question: trends — line; comparison — bar; relationship — scatter; distribution — histogram.
Topic 6
Predictive vs descriptive data mining
Predictive mining uses known data to predict an unknown or future value. Descriptive mining finds patterns that describe the data, with no target variable.
Goal
Predict unknown or future values
Describe patterns in existing data
Learning
Supervised — has a target variable
Unsupervised — no target
Techniques
Classification, regression, time-series prediction
Clustering, association rules, summarisation
Example
Will this customer churn?
What groups of customers exist?
Example
Telecom data: predicting which customers will leave next month is predictive (classification). Grouping customers by usage pattern is descriptive (clustering).
Exam tip
Classification and regression are predictive. Clustering, association and summarisation are descriptive.
Topic 7
Need for data analytics in business intelligence
BI and analytics work together: BI describes the business, while analytics diagnoses, predicts and recommends.
- BI reports show what happened; analytics and data mining explain why and predict what will happen, enabling proactive decisions.
- Applications: customer segmentation and targeting, churn prediction, credit scoring, fraud detection, market basket analysis, demand forecasting, predictive maintenance, HR attrition analysis.
- Challenges: data quality, privacy (DPDP Act 2023), skills, model bias, explaining results to managers.
Example
Moving beyond reports: a BI report shows that a bank's credit card defaults rose 3% last quarter. Analytics then shows which customer segments drove the rise and predicts which current customers are likely to default next.
Exam tip
Use the four levels: descriptive, diagnostic, predictive and prescriptive.
Key terms
- Data mining
- Discovering useful patterns in large data sets
- KDD
- Knowledge discovery in databases
- CRISP-DM
- Six-phase standard process for data mining
- Web usage mining
- Mining clickstream and log data
- Supervised learning
- Learning with a known target variable
Quick revision
- KDD steps; CRISP-DM six phases.
- Functionalities: characterisation, association, classification, prediction, clustering, outliers, trends.
- Text, web (content, structure, usage), sequence and context mining.
- Visualisation for exploration and evaluation.
- Predictive (supervised) vs descriptive (unsupervised); analytics in BI.
Important exam questions
Practice questions written to the PTU exam pattern for this unit's syllabus: short answers (Section A style) and long answers (Sections B and C style).
Short-answer questions
- Q1.Define data mining.
- Q2.Name the phases of CRISP-DM.
- Q3.What is text mining?
- Q4.Name the three types of web mining.
- Q5.Distinguish predictive and descriptive data mining.
- Q6.Give two business applications of data mining.
Long-answer questions
- Q1.Explain the data mining process using KDD and CRISP-DM.
- Q2.Explain the functionalities of data mining.
- Q3.Discuss text, web and sequence mining with examples.
- Q4.Explain the need for data analytics in business intelligence.
Stuck on this unit?
Message SBS on WhatsApp for help with Data Mining for Business Decisions, or to ask about studying MBA at Synetic.
