Google translated http://scikit-learn.org/0.18/tutorial/machine_learning_map/index.html Tutorial Table of Contents / Previous Tutorial

Choosing the right estimator

Often, the hardest part of solving a machine learning problem is finding the right estimator for the job. Different estimators are suitable for different types of data and problems. The flow chart below is intended to provide the user with a rough guide on which estimator to try the data with. Click the quote in the figure below to see the document. You can't click

scikit-learn algorithm cheat sheet

start

--Are there more than 50 samples? --NO → Let's collect more data

yes ↓ --Do you want to estimate the category? --Yes → Do you have labeled teacher data? --Yes → Classification --No → Clustering
- No ――Do you want to predict the amount? yes → regression -Just see? yes → Dimensionality reduction ――Do you want to predict the structure? ――It's not attached ...

Classification

--Is the sample less than 100,000?

Yes
- LinearSVC --It doesn't work --Text data? --yes → Naive Bayes
  - no -K-nearest neighbor method --It doesn't work ↓
    - SVC -Ensemble method
No -Stochastic Gradient Descent Classification --It doesn't work ↓ -Kernel approximation

Clustering

――Do you know how many categories there are?

I know --Less than 10,000 samples?
- Yes↓ -K-means clustering (K Means) --It doesn't work ↓ -Spectral Clustering -Mixed Gauss Model (GMM)
- no -Mini-batch k-means
Do not know --Less than 10,000 samples?
- Yes↓
  - MeanShift -Variational mixed Gaussian distribution --No → Not attached ...

Regression

--Is the sample less than 100,000? --no → Regression of stochastic gradient descent

yes ↓ --Are there few important features?
- yes
  - Elastic Net
  - Lasso
- no -Ridge Regression
  - SVR(kernel='linear') --These do not work well ↓ -Ensemble method
  - SVR(kernel='rbf')

Dimensionality reduction

Randomized PCA --It doesn't work ↓ --Less than 100,000 samples? --no → Kernel approximation
- yes ↓
  - Isomap -Spectrum embedding (Laplace proper map) --It doesn't work ↓
  - LLE

Tutorial Table of Contents / Next Tutorial