Unsupervised Machine Learning

Question 1

Unlabeled Data

Answer 1

any data that’s not organized in an easily identifiable manner is known as unstructured/unlabeled

Question 2

Goals of Unsupervised Learning

Answer 2

Goal is to learn about data’s underlying structure and find out how different features relate to each other.

Question 3

Name 2 Methodologies of unsupervised learning

Answer 3

1. Recommendation Systems
2. K-mean models

Question 4

Briefly describe a Recommendation system

Answer 4

Recommendation systems are a
- subclass of machine learning algorithms that
- can be both supervised or unsupervised
- offer relevant suggestions to users.

Question 5

What is the goal of a recommendation system?

Answer 5

to quantify how similar one thing is to another, and use this information to suggest a closely related option.

Question 6

What is content-based filtering?

Answer 6

Content-based filtering is a type of recommendation system where comparisons are made based on the attributes of the content itself.

For example, attributes of a song you played are compared to attributes of other songs to determine similarity.

Question 7

What are some benefits of content-based filtering?

Answer 7

• The benefits include being easy to understand, recommending more of what a user likes,
• not needing other users’ information to work, and - -
• being able to map users and items in the same space to recommend things that are closest to a user’s typical preferences.

Question 8

What are some drawbacks of content-based filtering?

Answer 8

1. Always recommends more of the same
2. Require manual input of attributes
3. Cannot reccommend across content type
4. Limited use cases

Question 9

What is collaborative filtering?

Answer 9

Collaborative filtering is a type of recommendation system that uses the likes and dislikes of users to make recommendations.

It does not need to know anything about the content itself. All that matters is if the user liked it.

Question 10

What are some benefits of collaborative filtering?

Answer 10

The benefits include the ability to
- recommend across content types,
- finding hidden correlations in the data, and
- not requiring tedious manual mapping.

Question 11

What are some drawbacks of collaborative filtering?

Answer 11

Drawbacks include
- needing lots of data to even start getting useful results,
- requiring every user to give the system lots of data, and
- dealing with sparse data that has a lot of missing values.

Question 12

What type of model is K-means and what does it do?

Answer 12

• unsupervised learning model
• partitioning algorithm,
• organize unlabeled data into clusters

Question 13

What is a Centroid?

Answer 13

• central point of a cluster
• also known as the mathematical mean

Question 14

List the 4 steps to build K-means model

Answer 14

1. Initiate k centroids
2. Assign all points to nearest centroid
3. Recalculate the centroid of each cluster.
4. Repeat Step 2 and 3 until the algorithm converges

Question 15

What is the difference between Clustering and Partitioning Algorithms

Answer 15

clustering algorithms: outlying points can exist outside of the clusters.

partitioning algorithms: all points must be assigned to a cluster.

in other words, K-means does not allow unassigned outliers

Question 16

What is k in Initiate k centroids step?

Answer 16

K = the number of centroids in your model, which is how many clusters you’ll have.

Question 17

Who makes decision on k = #?

Answer 17

you

Question 18

How to choose k value?

Answer 18

sometimes known, for instance, if there’s 3 species of beetle to cluster, then k=3. sometimes unknown.

Question 19

Name 2 other clustering methodologies

Answer 19

• DBSCAN (Density-Based Spatial Clustering of Applications with Noise): Groups data points together based on their density.
• Agglomerative clustering: Creates a hierarchy of clusters by merging data points or clusters iteratively

Question 20

What cluster shape does K-mean work best with?

Answer 20

round clusters

Question 21

DBSCAN (density-based spatial clustering of applications with noise)

Answer 21

• searches your data space for continuous regions of high density.
• Or find clusters based on density, the shape of the cluster isn’t as important as it is for K-means.

Question 22

DBSCAN Hyperparameters

Answer 22

epsilon, min_samples

Question 23

DBSCAN: eps, Epsilon (ε)

Answer 23

The radius of your search area from any given point

Question 24

DBSCAN: min_samples

Answer 24

the number of samples in an ε-neighborhood for a point to be considered a core point (including itself)

Question 25

Agglomerative clustering

Answer 25

works by first assigning every point to its own cluster, then progressively combining clusters based on intercluster distance.

Question 26

Agglomerative clustering requirement

Answer 26

you specify a desired number of clusters or a distance threshold, which is the linkage distance

Question 27

Agglomerative clustering: Linkage

Answer 27

different ways to measure the distances that determine whether or not to merge the clusters.

Question 28

Common Linkages

Answer 28

Single: The minimum pairwise distance between clusters. Complete: The maximum pairwise distance between clusters. Average: The distance between each cluster’s centroid and other clusters’ centroids. Ward: This is not a distance measurement. Instead, it merges the two clusters whose merging will result in the lowest inertia.

Question 29

When does Agglomerative clustering stop?

Answer 29

1. You reach a specified number of clusters.
2. You reach an intercluster distance threshold (clusters that are separated by more than this distance are too far from each other and will not be merged).

Question 30

Agglomerative clustering: Hyperaparameters

Answer 30

n_clusters: the number of clusters you want in your final model. linkage: the linkage method to use to determine which clusters to merge. affinity: the metric used to calculate the distance between clusters. Default = euclidean distance. distance_threshold: the distance above which clusters will not be merged

Question 31

Agglomerative clustering PROs

Answer 31

scales reasonably well, can detect clusters of various shapes.

Question 32

What is considered good clustering model?

Answer 32

1. clearly identifiable clusters (within each cluster or intracluster, the points are close to each other)
2. Each cluster is well separated from other clusters. (between the clusters themselves or intercluster, you want lots of empty space)

Question 33

K-means: metrics to evaluate good clusters

Answer 33

1. Inertia
2. Silhouette Score

Question 34

K-means: Inertia

Answer 34

Inertia is a metric used in K-Means clustering to measure the quality of the clustering.

It represents the average squared distance between each data point and its assigned cluster centroid

Lower Inertia, Better Clustering
The goal of K-Means is to minimize inertia.

A lower inertia indicates that the data points are more tightly clustered around their respective centroids, suggesting a better clustering solution.

Question 35

K-means: Silhouette Score

Answer 35

• more precise evaluation metric than inertia because it also takes into account the separation between clusters.

Silhouette score is defined as the mean of the silhouette coefficients of all the observations in the model.

Provides insight as to what the optimal value for K should be, and uses both intracluster and intercluster measurements in its calculation

Question 36

Inertia Score

Answer 36

• lower = better (less distance between each observation and its nearest centroid.
• 0 = useless (all points are overlapping each other in the center).

Question 37

Inertia Score PROs

Answer 37

• helps us to decide on the optimal k value.
• We do this by using the elbow method.

Question 38

Elbow Method

Answer 38

Plot of inertia vs K-values (1,2,3…etc).

• A good way of choosing an optimal k value is to find the elbow of the curve.
• This is the value of k at which the decrease in inertia starts to level off.

Question 39

Explain Silhouette Scores (-1,0,1)

Answer 39

1 = optimal (an observation sit nicely within its own cluster and well separated from other clusters).

0 = an observation is on the boundary between clusters.

-1 = in the wrong cluster.