Introduction to Machine Learning

In This Section, You Will Learn:

• • What Machine Learning actually is
• • Types of ML: Supervised vs Unsupervised
• • Common ML algorithms explained simply
• • Training, Testing, and Accuracy
• • Building your first prediction model
• • When to use ML (and when not to)

What is Machine Learning?

• Machine Learning = Teaching computers to learn patterns from data and make predictions.
• Think of it like:
• • You show a child 100 pictures of cats and dogs
• • The child learns what makes a cat vs a dog
• • Now the child can identify new cats and dogs they've never seen
• In Business Terms:
• • Show the model 1000 customers who churned
• • Model learns patterns of churning customers
• • Model predicts which current customers might churn
• The Magic: The model finds patterns humans can't see!

Types of Machine Learning

• 1. Supervised Learning (Most Common):
• • You have input data AND the correct answers
• • Model learns to map inputs to outputs
• • Example: Historical sales → Predict future sales
• • Types: Regression (numbers), Classification (categories)
• 2. Unsupervised Learning:
• • You have data but NO correct answers
• • Model finds hidden patterns itself
• • Example: Group customers by behavior (clustering)
• • Types: Clustering, Dimensionality Reduction
• 3. Reinforcement Learning:
• • Model learns by trial and error
• • Gets rewards for good actions, penalties for bad
• • Example: Game-playing AI, robotics
• Start with Supervised Learning. It's the most practical for business.

Common Algorithms Explained Simply

• Linear Regression:
• • Predicts a number (continuous value)
• • Example: Predict house price based on size
• • Use when: Target is a number
• Logistic Regression:
• • Predicts a category (Yes/No, Class A/B/C)
• • Example: Will customer buy? (Yes/No)
• • Use when: Target is a category
• Decision Trees:
• • Makes decisions like a flowchart
• • Easy to understand and explain
• • Example: If age > 30 AND income > 50k, likely buyer
• Random Forest:
• • Many decision trees voting together
• • More accurate than single tree
• • Used in many real-world applications

Training, Testing, and Accuracy

• The Process:
• 1. Split your data: 80% training, 20% testing
• 2. Train model on training data
• 3. Test model on testing data (unseen by model)
• 4. Measure accuracy
• Why Split the Data?
• • If you test on training data, model will cheat
• • Testing on new data shows true performance
• • This prevents 'overfitting'
• Accuracy Metrics:
• • Accuracy: % of correct predictions
• • Precision: Of predicted positives, how many were right?
• • Recall: Of actual positives, how many did we catch?
• • 80%+ accuracy is usually good, but depends on problem

Building Your First Model (Code)

• Using Scikit-Learn (The Standard Library):
• from sklearn.model_selection import train_test_split
• from sklearn.linear_model import LogisticRegression
• from sklearn.metrics import accuracy_score
• # Split data
• X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)
• # Create and train model
• model = LogisticRegression()
• model.fit(X_train, y_train)
• # Make predictions
• predictions = model.predict(X_test)
• # Check accuracy
• accuracy = accuracy_score(y_test, predictions)
• print(f'Accuracy: {accuracy:.2f}')
• Don't memorize this. Ask AI to write and explain it!

When to Use ML (And When Not To)

• Use ML When:
• • You have lots of historical data
• • There's a pattern to learn
• • Manual rules are too complex
• • You need to make predictions at scale
• Don't Use ML When:
• • Simple rules work fine
• • You don't have enough data
• • You can't explain the prediction (sometimes important)
• • The problem changes too often
• Example:
• • ML Good: Predicting which of 1M customers will churn
• • ML Overkill: Deciding if someone is an adult (just check age >= 18)