Learning Objectives
By the end of this lesson, the learner should be able to:
- Define machine learning and explain its fundamental principles.
- Explain the relationship between machine learning, artificial intelligence, data analytics and statistics.
- Distinguish machine learning from traditional rule-based programming.
- Explain how machine learning systems learn from data.
- Identify the major components of a machine learning problem.
- Distinguish between training data, validation data and test data.
- Explain the importance of features, labels and observations.
- Describe common machine learning tasks in business.
- Explain the machine learning workflow at a conceptual level.
- Identify the opportunities and limitations of machine learning in business environments.
1.1 Understanding Machine Learning
Machine learning is a field of computing and analytical science concerned with developing systems that can learn patterns from data and use those patterns to make predictions, classifications, recommendations or decisions without requiring every decision rule to be explicitly programmed.
In conventional programming, a programmer typically specifies rules that transform inputs into outputs:
Input + Rules → Output
For example, a payroll system might contain explicitly programmed rules such as:
If an employee works more than 40 hours, calculate overtime according to the applicable overtime rate.
The programmer determines the rules in advance.
Machine learning takes a different approach. Rather than explicitly specifying every rule, the system is provided with data from which it can identify patterns:
Data + Learning Algorithm → Model
The resulting model can then be applied to new data:
New Data + Trained Model → Prediction or Decision
For example, instead of manually defining every rule for identifying potentially fraudulent transactions, a financial institution may train a machine learning model using historical transaction data containing examples of legitimate and fraudulent activity.
The model may identify relationships involving:
- Transaction amount
- Transaction location
- Transaction frequency
- Time of transaction
- Merchant characteristics
- Customer behavior
- Device information
- Previous transaction patterns
The model can then estimate whether a new transaction is likely to be fraudulent.
1.2 Why Machine Learning Matters in Business
Organizations have traditionally relied on human judgment, predefined business rules and descriptive reporting to make decisions.
However, modern organizations increasingly deal with:
- Large volumes of data
- High-frequency transactions
- Rapidly changing customer behavior
- Complex relationships between variables
- Real-time decision requirements
- Increasingly digital business processes
These conditions can make purely manual analysis difficult.
Machine learning can help organizations analyze large datasets and identify patterns that may not be immediately obvious to human analysts.
For example, an e-commerce company could use machine learning to estimate which customers are most likely to purchase a particular product.
A bank could use machine learning to estimate the probability that a borrower will default.
A telecommunications company could predict which customers are likely to leave for a competitor.
A manufacturing company could predict when equipment is likely to fail.
The underlying objective is not simply to “use AI.” The objective is to solve a meaningful business problem using data-driven methods.
1.3 Machine Learning, Artificial Intelligence and Business Analytics
These concepts are closely related but should not be treated as identical.
Artificial Intelligence
Artificial intelligence (AI) is the broader field concerned with developing systems capable of performing tasks that traditionally require aspects of human intelligence.
Examples include:
- Reasoning
- Perception
- Language processing
- Decision-making
- Planning
- Learning
Machine Learning
Machine learning is one important approach within artificial intelligence.
A simplified relationship is:
Artificial Intelligence → Machine Learning → Specific Learning Methods and Models
Machine learning enables systems to learn patterns from data rather than relying exclusively on manually programmed rules.
Business Analytics
Business analytics focuses on using data, analytical methods and technology to support business decisions and improve organizational performance.
Business analytics may include:
- Descriptive analytics
- Diagnostic analytics
- Predictive analytics
- Prescriptive analytics
- Statistical analysis
- Data visualization
- Machine learning
Therefore, machine learning is a tool within the broader business analytics ecosystem, rather than a replacement for business analytics.
1.4 Machine Learning and Traditional Statistical Analysis
Machine learning and statistics overlap considerably.
Both can be used to:
- Identify relationships in data
- Make predictions
- Quantify uncertainty
- Analyze patterns
- Support decision-making
However, their emphasis can differ.
Traditional statistical analysis often places substantial emphasis on:
- Inference
- Estimation
- Hypothesis testing
- Understanding relationships between variables
- Quantifying uncertainty
- Explaining phenomena
Machine learning often places greater emphasis on:
- Prediction
- Generalization to new data
- Automated pattern detection
- Model performance
- Handling large and complex datasets
The distinction should not be overstated. Modern data science frequently combines statistical reasoning with machine learning techniques.
A business analyst should therefore understand both predictive performance and the business meaning of the relationships being modeled.
1.5 Rule-Based Programming versus Machine Learning
Consider a bank attempting to identify potentially risky loan applications.
Traditional rule-based approach
The bank might establish rules such as:
- If income is below a certain threshold → high risk.
- If credit score is below a specified level → high risk.
- If debt-to-income ratio exceeds a threshold → high risk.
These rules are explicitly defined.
Machine learning approach
Instead, the bank could provide historical loan data containing:
- Applicant characteristics
- Loan information
- Credit history
- Repayment behavior
- Default outcomes
The machine learning algorithm identifies patterns associated with different outcomes.
The resulting model can then evaluate a new applicant.
The machine learning approach does not necessarily eliminate business rules. In real organizations, business rules and machine learning models can operate together.
1.6 What Does It Mean for a Machine to “Learn”?
The term “learning” can be misleading.
A machine learning system does not learn in exactly the same way that a human learns.
Instead, the system uses an algorithm to identify patterns or relationships within data and represents those patterns through a mathematical model.
A simplified process is:
Historical Data → Algorithm → Model → New Data → Prediction
Suppose a retailer has five years of sales information.
The dataset might contain:
|
Date |
Product |
Price |
Promotion |
Sales |
|
Jan 1 |
A |
100 |
Yes |
250 |
|
Jan 2 |
A |
100 |
No |
140 |
|
Jan 3 |
A |
95 |
Yes |
280 |
|
… |
… |
… |
… |
… |
A machine learning algorithm can analyze relationships among the variables.
It may discover that:
- Promotional periods tend to increase sales.
- Price reductions are associated with higher demand.
- Demand varies by season.
- Certain products respond more strongly to promotions.
The model can then use these relationships to estimate future sales.
1.7 Core Components of a Machine Learning Problem
A basic machine learning problem usually involves several important components.
1. Data
Data provides the information from which patterns are learned.
Examples include:
- Customer transactions
- Financial records
- Website activity
- Sales records
- Sensor readings
- Employee records
- Survey responses
2. Features
Features are the variables used by a machine learning model to identify patterns or make predictions.
For example, when predicting customer churn, features might include:
- Customer age
- Monthly spending
- Number of complaints
- Contract duration
- Number of service interruptions
- Usage frequency
3. Target or Label
For supervised learning problems, the target represents the outcome the model is trying to predict.
For example:
Customer churn prediction
Target:
Churn = Yes/No
Sales prediction
Target:
Expected monthly sales
4. Algorithm
An algorithm is the computational procedure used to learn patterns from the data.
Examples include:
- Linear regression
- Logistic regression
- Decision trees
- Random forests
- Support vector machines
- Neural networks
- K-means clustering
These methods will be examined more extensively in later lessons.
5. Model
The model is the learned representation of patterns in the data.
It can subsequently be used to generate predictions or other analytical outputs.
1.8 Observations, Features and Targets
Understanding the structure of a dataset is fundamental to machine learning.
Suppose a company has data about 10,000 customers.
Each row may represent one customer.
Each column may represent a variable.
For example:
|
Customer |
Age |
Monthly Spend |
Complaints |
Churn |
|
C001 |
24 |
3,500 |
1 |
No |
|
C002 |
41 |
7,200 |
4 |
Yes |
|
C003 |
29 |
5,100 |
0 |
No |
Here:
- Each row is an observation.
- Age, Monthly Spend and Complaints are features.
- Churn is the target/label.
The distinction becomes particularly important when preparing data for machine learning.
1.9 Training, Validation and Test Data
A model should not simply be evaluated using the same data from which it learned.
Doing so can create a misleading impression of performance.
Machine learning datasets are therefore commonly divided into separate subsets.
Training data
Training data is used to teach the model.
The algorithm examines relationships within this data and adjusts the model accordingly.
Validation data
Validation data can be used during model development to:
- Compare models
- Tune parameters
- Select approaches
- Identify potential overfitting
Test data
Test data is held aside and used to provide an independent assessment of how well the final model performs on previously unseen data.
A simplified structure is:
Dataset
→ Training Set
→ Validation Set
→ Test Set
Not every machine learning workflow uses all three datasets in exactly the same way, but the fundamental principle is important:
A model should demonstrate that it can generalize beyond the data used to train it.
1.10 The Concept of Generalization
Generalization refers to a model’s ability to perform effectively on new, unseen data.
Consider a model trained to identify fraudulent transactions.
If the model performs extremely well on historical transactions but performs poorly on new transactions, it has limited practical value.
The goal is therefore not simply:
“Build a model that performs well on the training data.”
The goal is:
“Build a model that learns useful patterns that remain effective when applied to new business data.”
This principle is central to machine learning.
1.11 Overfitting and Underfitting
Two common problems are overfitting and underfitting.
Overfitting
Overfitting occurs when a model learns the training data too closely, including patterns that do not generalize well to new data.
For example, a model might achieve 99% accuracy on training data but only 70% on unseen data.
This suggests that the model may have memorized aspects of the training data rather than learning generalizable patterns.
Underfitting
Underfitting occurs when a model is too simple to capture important patterns in the data.
For example, a model might perform poorly on both training and test data because it fails to represent the underlying relationships adequately.
The objective is to develop a model that achieves an appropriate balance between learning meaningful patterns and generalizing to new observations.
1.12 Common Business Applications of Machine Learning
Machine learning is used across many business functions.
Customer Analytics
Organizations can use machine learning to:
- Predict customer churn
- Segment customers
- Recommend products
- Estimate customer lifetime value
- Identify purchasing patterns
Financial Services
Applications include:
- Credit-risk prediction
- Fraud detection
- Transaction monitoring
- Loan default prediction
- Risk assessment
Marketing
Machine learning can support:
- Campaign targeting
- Customer propensity modeling
- Recommendation systems
- Advertising optimization
- Lead scoring
Operations
Organizations can use machine learning for:
- Demand forecasting
- Inventory optimization
- Predictive maintenance
- Supply-chain analytics
- Resource planning
Human Resources
Potential applications include:
- Workforce forecasting
- Employee-retention analysis
- Recruitment analytics
- Skills analysis
However, HR applications require particularly careful attention to fairness, privacy and discrimination risks.
1.13 Machine Learning Workflow
A simplified machine learning workflow can be represented as:
- Define the Business Problem
↓
- Collect Relevant Data
↓
- Prepare and Clean Data
↓
- Select Features
↓
- Split Data
↓
- Select a Machine Learning Approach
↓
- Train the Model
↓
- Evaluate the Model
↓
- Interpret Results
↓
- Deploy and Monitor
This demonstrates an important principle:
Machine learning begins with a business problem—not with an algorithm.
An organization should not adopt machine learning simply because it is technologically fashionable.
The organization should first determine:
- What decision needs improvement?
- What outcome needs to be predicted?
- What data is available?
- What business value could result?
- What risks are involved?
1.14 Business Value versus Technical Performance
A machine learning model can have excellent technical performance and still fail to create business value.
For example, suppose a bank develops a highly accurate model for predicting loan defaults.
However:
- The model takes several hours to produce predictions.
- Loan officers need decisions within seconds.
- The model is difficult to interpret.
- The required data is not available in real time.
Despite strong predictive performance, the model may have limited operational value.
Business analytics professionals must therefore consider both:
Technical performance
and
Business usefulness
A successful machine learning initiative should ideally contribute to measurable outcomes such as:
- Increased revenue
- Reduced costs
- Reduced fraud
- Improved customer retention
- Faster decision-making
- Improved operational efficiency
- Reduced risk
1.15 Limitations of Machine Learning
Machine learning is powerful, but it is not universally appropriate.
Important limitations include:
Poor-quality data
A model cannot automatically transform fundamentally unreliable data into reliable information.
Historical bias
If historical data contains systematic bias, a model may reproduce or amplify that bias.
Lack of explainability
Some complex models can be difficult for decision-makers to interpret.
Changing business environments
Patterns learned from historical data may become less useful when market conditions change.
Data privacy concerns
Sensitive information must be handled according to applicable laws, regulations and organizational policies.
Implementation costs
Machine learning may require:
- Skilled personnel
- Data infrastructure
- Computing resources
- Integration with existing systems
- Ongoing monitoring
Therefore, machine learning should be approached as a business capability, not merely a software installation.
1.16 The Role of the Business Analyst
The business analyst plays an important role in translating organizational problems into analytical questions.
For example:
Business problem:
Customer retention is declining.
Analytical question:
Which customers are most likely to leave within the next 90 days?
Potential machine learning problem:
Develop a classification model that estimates the probability of customer churn.
Business action:
Target high-risk customers with appropriate retention interventions.
This illustrates the connection between business analytics and machine learning:
Business Problem → Analytical Question → Data → Model → Insight → Business Action
The model itself is not the final objective.
The business decision is.
1.17 Practical Business Example: Customer Churn
Consider a telecommunications company with 500,000 customers.
Management wants to reduce customer churn.
The company has historical data containing:
- Customer tenure
- Monthly expenditure
- Number of complaints
- Data usage
- Call usage
- Contract type
- Payment behavior
- Service interruptions
- Previous churn status
The organization can use historical records to train a model to estimate the probability that a customer will leave.
The analytical process could be:
Historical customer data
↓
Feature preparation
↓
Machine learning model
↓
Churn probability
↓
Customer prioritization
↓
Retention intervention
The business team can then determine which interventions are appropriate.
This example demonstrates that machine learning is most valuable when it connects analytical predictions to specific managerial actions.
1.18 Key Principles for Business Application
When considering machine learning, organizations should ask:
- Is there a clearly defined business problem?
- Is machine learning actually appropriate for the problem?
- Is sufficient data available?
- Is the data reliable?
- Is the target outcome clearly defined?
- Can model performance be measured?
- Can the results be interpreted appropriately?
- Can the model be integrated into business processes?
- What risks could the model introduce?
- How will the model be monitored after deployment?
These questions help prevent organizations from implementing machine learning simply because it is technologically attractive.
Reflection Exercise
Consider a retail business experiencing declining sales.
Management wants to use machine learning to improve performance.
Reflect on the following questions:
- What specific business problem should management define before selecting an algorithm?
- What data could potentially be used?
- What could be the target variable?
- What features might help predict the target?
- How could the resulting predictions support management decisions?
- What risks might arise if the historical data is biased or incomplete?
Best Practices
Business analytics professionals should observe the following practices when introducing machine learning:
- Start with the business problem, not the technology.
- Understand the data before selecting a model.
- Separate training and evaluation data appropriately.
- Focus on generalization, not merely training performance.
- Measure both technical and business performance.
- Document assumptions and limitations.
- Consider ethical and privacy implications.
- Ensure appropriate human oversight for consequential decisions.
- Monitor models after deployment.
- Review models periodically as business conditions and data change.
Lesson Summary
Machine learning enables organizations to use data to identify patterns and generate predictions, classifications, recommendations and other analytical outputs.
The lesson established that machine learning is closely related to, but distinct from, artificial intelligence, statistics and business analytics. Machine learning differs from conventional rule-based programming because the model learns patterns from data rather than relying entirely on manually specified rules.
Key concepts introduced include:
- Data
- Features
- Targets and labels
- Algorithms
- Models
- Training data
- Validation data
- Test data
- Generalization
- Overfitting
- Underfitting
The lesson also demonstrated that machine learning can support business functions including finance, marketing, operations, customer analytics and risk management.
Most importantly, machine learning should be viewed as a means of addressing business problems and improving decisions, rather than as an end in itself.
The fundamental relationship is:
Business Problem → Data → Machine Learning Model → Prediction/Insight → Business Decision → Business Value
References
- IBM — Machine Learning
IBM Machine Learning overview - Google for Developers — Introduction to Machine Learning
Google Introduction to Machine Learning - Microsoft Learn — Machine Learning Fundamentals
Microsoft Learn: Machine Learning - scikit-learn — Machine Learning Documentation
scikit-learn User Guide - OECD — Artificial Intelligence and Data
OECD Artificial Intelligence Policy Observatory