The document discusses supervised and unsupervised machine learning techniques, focusing on unsupervised models like K-means clustering and hierarchical clustering, as well as the Random Forest algorithm. Random Forest is an ensemble learning method that utilizes bagging and feature bagging to create a more accurate model through multiple decision trees. It highlights the advantages, such as versatility and ease of understanding, alongside disadvantages like increased model complexity with more trees.

1. SUPERVISED AND
UNSUPERVISED LEARNING
TECHNIQUES
-Ms. Keren Evangeline. I / AP(O.G)
SRM Valliammai Engineering College
Department of Medical Electronics
Faculty Development Programme

14. LIST OF COMMON UNSUPERVISED ML
MODELS
• K – Means Clustering
• Principal component analysis (Dimensionality
Reduction)
• Hierarchial clustering

15. ADVANTAGES OF UNSUPERVISED ML MODELS

16. DISADVANTAGES OF UNSUPERVISED ML MODELS

17. DECISION TREE

24. [Size < 2000?]
/
Yes (<=2000) No (>2000)
/
[Bedrooms < 3?] [Lot Size < 5000 sq.ft?]
/ /
Yes (<=3) No (>3) Yes (<5000) No (>5000)
/ /
[Year < 1990?] [Price: $400K] [Price: $450K] [Price:$550K]
/
Yes No
/
[Price: $300K] [Price: $350K]
Continuous Variable Decision Tree

32. RANDOM
FOREST

33. Introduction
• A random Forest Algorithm is a supervised machine learning Algorithm consisting
decision trees.
• The general method of random decision forests was first proposed by Ho in 1995.
After that, It was developed by Leo Breiman in 2001.
R A N D O M F O R E S T 33

34. Decision Tree
R A N D O M F O R E S T 34

35. Decision Tree – Important Terms
R A N D O M F O R E S T 35

36. How Decision Tree Works
• It follows a tree-like model of decisions and their possible consequences.
• The algorithm works by recursively splitting the data into subsets based on the most significant feature at each
node of the tree.
R A N D O M F O R E S T 36

37. How Decision Tree Works
R A N D O M F O R E S T 37

38. Ensemble Learning
Ensemble learning creates a stronger model by aggregating the predictions of multiple weak models. Random
Forest is an example of ensemble learning where each model is a decision tree. The idea behind it is- the
wisdom of the crowd. The majority vote aggregation can have better accuracy than the individual models.
R A N D O M F O R E S T 38

39. Dataset
Preparation
Bootstrap Aggregating (Bagging)
Creating a different training subset randomly from the
original training dataset with replacement is called
Bagging. With replacement refers to the bootstrap sample
having duplicate elements. Reduces variance, helps to
avoid overfitting.
Out Of Bag (OOB)
The out-of-bag dataset represents the remaining elements
which were not in the bootstrap dataset. Used for cross
validation or to evaluate the performance.
Random Subspace Method(Feature Bagging)
While building the tree, for splitting, a randomly selected
subset of the features are used. So, the trees are
more different having low correlation.
R A N D O M F O R E S T 39

40. Bagging at training time​
R A N D O M F O R E S T 40

41. Bagging at inference time
R A N D O M F O R E S T 41

42. Random Subspace Method at training time
R A N D O M F O R E S T 42

43. Random Subspace Method at inference time
R A N D O M F O R E S T 43

44. Definition
The Random Forest algorithm is an ensemble learning method consisting of many decision
trees that are built by using bagging and feature bagging which helps to create an
uncorrelated forest of trees whose combined prediction is more accurate than that of a single
tree. For classification tasks, final prediction is done by taking majority votes and for
regression tasks, average of all the individual trees.
R A N D O M F O R E S T 44

45. Random Forest Model
R A N D O M F O R E S T 45

46. Why Use Random Forest
 Random forests are an effective tool in prediction.
 Forests give results competitive with boosting and adaptive bagging, yet do not
progressively change the training set.
 Random inputs and random features produce good results in classification- less so in
regression.
 For larger data sets, we can gain accuracy by combining random features with boosting.
R A N D O M F O R E S T 46

47. Advantages
 Versatile uses
 Easy-to-understand hyperparameters
 Classifier doesn't overfit with enough
trees
Disadvantages
 Increased accuracy requires more trees
 More trees slow down model
 Can’t describe relationships within data
Advantages and Disadvantages

48. Random Forest Applications
 Detects reliable debtors and potential fraudsters in finance
 Verifies medicine components and patient data in healthcare
 Gauges whether customers will like products in e-commerce
R A N D O M F O R E S T 48

49. Thank you
Any Question?
R A N D O M F O R E S T 49