Selecting the Right ML Techniques for Use Cases (Based on AWS Best Practices)
Machine learning problems are typically categorized as supervised (classification and regression), unsupervised (clustering, anomaly detection), and can involve techniques like linear/logistic regression or cluster analysis. Supervised learning uses labeled data to predict outcomes, while unsupervised learning discovers patterns in unlabeled data.
Does Your Dataset Contain Target Values?
- If Yes → Use Supervised Learning
- If No → Use Unsupervised Learning
Supervised Learning
Supervised learning applies when your dataset contains target values (labels) you want to predict. The type of values you want to predict determines the technique:
- Categorical Target Values:
Use Classification techniques.- Example: Predicting whether an email is spam or not spam.
- Continuous Target Values:
Use Regression techniques.- Example: Predicting the price of a house given its features.
Unsupervised Learning
Unsupervised learning is used when your dataset does not contain labeled target values. Instead, the goal is to discover patterns or structure within the data:
- Discrete Groups:
Use Clustering techniques.- Example: Segmenting customers into groups based on purchasing behavior.
- Outliers (Anomalies):
Use Anomaly Detection techniques.- Example: Detecting unusual credit card transactions.
Summary Table
| Dataset Contains Target? | Learning Type | Value Type | Technique |
|---|---|---|---|
| Yes | Supervised Learning | Categorical | Classification |
| Yes | Supervised Learning | Continuous | Regression |
| No | Unsupervised Learning | Discrete Groups | Clustering |
| No | Unsupervised Learning | Outliers | Anomaly Detection |
In summary, ML techniques start with identifying whether the dataset has target labels. If so, supervised learning (classification or regression) is used. If not, unsupervised learning (clustering or anomaly detection) is appropriate, depending on whether you want to group data or find outliers.