Supervised Machine Learning
Dr. Philip Jiang
School of Management and Marketing Operation | WMS
Supervised Learning
Data Label
Car
Wheels, Wings, Colors, Legs
Car
Car Feature Extraction Prediction Model
Insect
Insect
Insect
Insect
Unsupervised Learning
Data Features
Wheels, Windows
Wheels, Windows
Wheels, Windows Commonality Clustering
Wings, Legs
Wings, Legs
Wings, Legs
Supervised Learning Unsupervised Learning
Data Type Labelled data (input-output pairs) Unlabeled data
Main Goal Predict outcomes for new data Extract hidden patterns or insights from data
Classification (e.g., spam detection), Regression Clustering (e.g., K-means), Association (e.g.,
Techniques
(e.g., sales prediction) market basket analysis)
Spam detection, sentiment analysis, pricing Anomaly detection, recommendation engines,
Applications
predictions, weather forecasting customer segmentation, medical imaging
Relatively simple; can be implemented with tools More computationally complex; requires handling
Complexity
like R or Python large, unclassified datasets
Human Requires human labeling of data; time-consuming Requires human validation of output to ensure
Intervention and requires expertise meaningful insights
May result in less accurate outcomes unless
Accuracy High accuracy due to labeled data
validated by humans
Risk of inaccurate results without human
Time-consuming to label data, can be labour-
Drawbacks validation, lacks transparency in how data is
intensive
clustered
[Link]
In model form
[Link]
K-Nearest Neighbor
Decision Tree Analysis
Random Forest
Supervised
Algorthims
Linear Regression
Logistic Regression
K-Nearest Neighbours
“Birds of a feather flock together.”
[Link]
White Wine or Red Wine
Red or White
[Link]
Decision Tree (Divide & Conquer)
Nodes
Edges/Branch
Leaf nodes
[Link]
Basic Divide-and-Conquer
Algorithm
Select a test for root node. Split instances into
Create branch for each subsets. One for each
possible outcome of the branch extending from the
test. node.
Repeat recursively for
Stop recursion for a
each branch, using only
branch if all its instances
instances that reach the
have the same class.
branch.
[Link]
From Decision tree to Random
Forest
Random Forest
is Average of
Trees
• [Link]
tree-vs-random-forest-algorithm/
Logistic vs linear
Regression
Linear Regression
Ypred = ax+ b
Cost Function
a
Gradient
descent
b
Linear Regression Cost Function
Logistic Regression
Linear vs Logistic Regression
What should be the price of the car Whether to buy car or not?
based on features? Whether to give loan or not?
How much loan can be given and at
what interested rate?
Linear vs logistic Regression
[Link]
THANK YOU