Machine Learning Practices

IIT Madras

Introduction to

Scikit-Learn (sklearn)

• Consistency
• Inspection
• Nonproliferation of classes
• Composition
• Sensible defaults

Introduction

• Estimators
• Predictors
• Transformers

Types of sklearn objects

• Estimates model parameter
• fit( )
• Examples- Imputer, LinearRegression

Estimators

• Some estimators are capable of making predictions on a given dataset
• predict( )
• score( )
• Examples-  LinearRegression

Predictors

• Some estimators can transform datasets
• transform( )
• fit_transform( )

Transformers

• Some estimators can transform datasets
• Can be arranged in the following manner:

1. Data 
2. Models
3. Model evaluation
4. Model inspection and selection

sklearn API

Supervised learning models-

1. Regression
2. Classification

​Unsupervised learning models-

Clustering

Models

Regression classes include:

• Classicial linear regression models
• Linear regression 
• Bayesian linear regression
• Outlier robust regression
• Multi-task linear regression models
• Generalized linear models of regressions
• sklearn.trees 
• sklearn.multioutput 

Regression

• sklearn.linear_model - classical algorithms like logistic regression, SVM etc.
• sklearn.svm - Support Vector Machine algorithms.
• sklearn.trees - decision tree-based models for classification and regression.
• sklearn.neighbors -  k-nearest neighbors algorithm.
• sklearn.naive_bayes- Naive-bayes classification.
• sklearn.multiclass - multi-class classification models.
• sklearn.multioutput - multioutput classification and regression.

Classification

• sklearn.metrics -  different metrics for model evaluation.
  1. Classification metrics
  2. Regression metrics
  3. Clustering metrics

Model evaluation

• Model selection module: sklearn.model_selection
• Model inspection: sklearn.inspection

Model inspection and selection