Dimensionality Reduction in Python
| Course Details | ||||||||
|---|---|---|---|---|---|---|---|---|
|
||||||||
| Languages | ||||||||
|
||||||||
| Outcomes | ||||||||
|
||||||||
| Price & Terms | ||||||||
|
||||||||
| Technical | ||||||||
|
||||||||
| Specification | ||||||||
|
||||||||
Learn to reduce dimensionality in Python with t-SNE, variance and correlation filters, model-based feature selection and PCA.
Datasets with many dimensions are hard to explore and expose a model to the curse of dimensionality. Here you work in Python to shrink them. You spot columns that carry no variance, are mostly missing or duplicate others. You let models rank which features matter for a prediction, then compress what remains into uncorrelated principal components. Plots made with t-SNE help you see the structure of the data before any trimming begins.
What You Will Achieve
- Fit t-SNE to a high-dimensional dataset and read the resulting visualization
- Drop columns with low variance, many missing values or strong pairwise correlation, choosing sensible thresholds
- Rank and keep predictive features with recursive feature elimination, random forests and LASSO, then combine their votes
- Add PCA to a model pipeline and decide how many components to keep based on explained variance
Program Overview
Exploring and Visualizing High-Dimensional Data
Why dimensionality reduction matters, how selection and extraction differ, and how to look at data that has many dimensions.
- Counting dimensions and removing features without variance
- Feature selection versus feature extraction
- Visually detecting redundant features
- t-SNE visualization of the ANSUR data
Feature Selection for Information and Model Accuracy
Two chapters on cutting features: first those that add little information, then those a model finds unhelpful for predicting a target.
- The curse of dimensionality and train-test evaluation
- Variance thresholds, missing values and pairwise correlation
- Recursive feature elimination and random forest importance
- LASSO regularization and combining several feature selectors
Feature Extraction with Principal Component Analysis
A close look at PCA, applied to data exploration, to preprocessing in a modeling pipeline and to compressing images.
- Manual feature extraction and principal component intuition
- Explained variance and choosing the number of components
- PCA in a model pipeline
- PCA for image compression
Certificate Awarded (Paid)
Complete the course to earn it.
Job Outlook
Feature selection and PCA are used in data science and machine learning work to cut complexity before building predictive models and to explore wide datasets.
Good for: Python users working with wide tabular datasets who want hands-on practice deciding which features to keep, drop or combine.
Keep in mind: t-SNE is used here only to visualize high-dimensional data; PCA is the one extraction method taken into a modeling pipeline.
Bottom line: Its strength is the side-by-side practice with variance, correlation and model-based selectors, closing with a PCA chapter that ties extraction to a modeling pipeline.
Frequently Asked Questions
What is the difference between feature selection and feature extraction?
Feature selection keeps a subset of the original columns and drops the rest, such as those with low variance or little predictive value. Feature extraction calculates new features, like uncorrelated principal components, that combine information from the originals. The course practices both.
Which examples do the exercises work on?
The t-SNE lessons fit the ANSUR data, the model-based selection lessons build a diabetes classifier, and the PCA chapter ends with an image compression use case.
How does the course combine several feature selectors?
The final lessons of the second feature selection chapter have you build a LassoCV regressor and other models, treat each one as a vote on which features matter, and combine three selectors to decide which features are worth keeping.
How do I choose how many principal components to keep?
The PCA chapter looks at how much variance each component explains, then has you select the proportion of variance to keep and choose the number of components from it, before using PCA in a model pipeline.
User Reviews
Write your review
You must be logged in to post a review.
Reviews can only be posted by registered learners. Log in or create a free account, then you will return here to post.
There are no reviews yet.