Semi-Supervised and Unsupervised Machine Learning Algorithms

Name: Semi-Supervised and Unsupervised Machine Learning Algorithms
Author: Briana Brownell

Discover the differences between and applications for semi-supervised and unsupervised machine learning

Preview the first lesson free — get full access to all 5 lessons.

Course: On-Demand

In supervised learning, we're trying to predict an outcome variable, but in unsupervised learning, there is no outcome variable. The beauty and difficulty of unsupervised machine learning techniques is that they find emergent patterns from the dataset itself. When doing supervised machine learning, data collection is important, but with unsupervised machine learning, it's even more important. It's recommended to be careful in your data collection to be sure the results will generalize. An emergent pattern may be an artifact of the dataset itself, and not generalizable. In that case, the results of an unsupervised machine learning model would be not only useless, but quite misleading. That's why the best practice is to take special care to assess the stability of unsupervised machine learning models that we build. There are two overarching styles of unsupervised machine learning, clustering and dimension reduction. Clustering algorithms group rows of data together. For example, if your dataset is the result of a survey, then it will group survey respondents together into clusters that are similar to each other, and different from respondents in other clusters. This is useful in marketing because marketers can understand and target specific customer segments with products designed according to their unique needs. The second style of unsupervised machine learning is dimension reduction. Here, instead of grouping rows of data together, we're grouping the columns. In other words, dimension reduction looks at the relationships between variables and groups similar variables together. A good example of dimension reduction is in artistic taste, such as music, movies, and books. If you ask a group of people how much they like a series of songs and artists, the relationship between those variables creates genres. For instance, someone who enjoys the rolling stones is likely to also enjoy other classic rock acts. Incidentally, this is how companies used to decide what to play on new radio stations to set up in a specific city. They could survey the musical taste of residents and determine what stations could be established, thus adding credibility to a business plan for a new station and its application for a broadcast license. Unsupervised learning is also why you see the same styles of stations in similar size markets and why larger cities have more genre specific radio stations.

Intermediate Provider : Briana Brownell 5 Lessons · 12m in English

Course Description

Unsupervised machine learning allows a practitioner to surface emergent patterns in data. These emergent patterns can give a machine learning practitioner important information about the structure of the data that can be used in many different applications.

In these lessons, we will discuss two important applications of unsupervised machine learning that showcase the two overarching styles of unsupervised machine learning: clustering, which groups data cases together, and dimension reduction, which groups data variables together. Clustering can be used in a customer segmentation which attempts to find groups of customers with similar behaviours, attitudes, and psychographic. Products can then be developed to specifically target those groups. Dimension reduction can be used to define the optimal genre for a new radio station by looking at similarities between musical styles that could be grouped together. These lessons will also discuss how unsupervised machine learning can be used in feature engineering. You’ll also explore semi-supervised machine learning and its applications.

What You'll Learn

Differentiate between problems that can be solved by clustering and dimension reduction
Determine data-collection methods that yield appropriate data for clustering and dimension reduction
Apply best practices in design decisions to minimize issues with unsupervised machine learning models
Use clustering to group data cases together, such as in customer segmentation
Use dimension reduction to group data variables together based on similarities
Explore how unsupervised machine learning supports feature engineering and semi-supervised learning

Key Takeaways

Unsupervised machine learning surfaces emergent patterns that reveal the structure of data for use in many applications.
Clustering and dimension reduction are the two overarching styles of unsupervised machine learning, grouping data cases and data variables respectively.
Clustering can power customer segmentation by finding groups of customers with similar behaviours, attitudes, and psychographics so products can target those groups.
Dimension reduction can identify an optimal radio-station genre by grouping similar musical styles together.
Unsupervised machine learning can be applied to feature engineering, and semi-supervised machine learning has its own applications.

Frequently Asked Questions

What does this course cover?

It covers two important applications of unsupervised machine learning representing its two overarching styles: clustering, which groups data cases together, and dimension reduction, which groups data variables together. It also covers using unsupervised learning for feature engineering and explores semi-supervised machine learning and its applications.

What are some real-world applications discussed in the course?

The course discusses customer segmentation through clustering to find groups of customers with similar behaviours, attitudes, and psychographics, and using dimension reduction to define the optimal genre for a new radio station by grouping similar musical styles.

What skills will I gain from this course?

The course is associated with skills in feature learning, machine learning, machine learning algorithms, machine learning methods, supervised learning, and unsupervised learning.

What lessons are included?

The course includes lessons on Unsupervised Machine Learning, Clustering, Dimension Reduction, Unsupervised Learning for Feature Engineering, and Semi-Supervised Machine Learning.

What will I be able to do after completing the course?

You will be able to differentiate between problems solvable by clustering versus dimension reduction, determine appropriate data-collection methods for these methods, and apply best practices in design decisions to minimize issues with the unsupervised machine learning model.