Machine learning with scikit-learn

1. Machine learning with scikit-learn

Hi, and welcome! My name is George Boorman, and I'll be your instructor for this course on supervised learning with scikit-learn.

2. What is machine learning?

Machine learning is the process whereby computers learn to make decisions from data without being explicitly programmed.

3. Examples of machine learning

For example, learning to predict whether an email is spam or not spam given its content and sender. Or learning to cluster books into different categories based on the words they contain, then assigning any new book to one of the existing clusters.

4. Unsupervised learning

Unsupervised learning is the process of uncovering hidden patterns and structures from unlabeled data. For example, a business may wish to group its customers into distinct categories based on their purchasing behavior without knowing in advance what these categories are. This is known as clustering, one branch of unsupervised learning.

5. Supervised learning

Supervised learning is a type of machine learning where the values to be predicted are already known, and a model is built with the aim of accurately predicting values of previously unseen data. Supervised learning uses features to predict the value of a target variable, such as predicting a basketball player's position based on their points per game. This course will exclusively focus on supervised learning.

6. Types of supervised learning

There are two types of supervised learning. Classification is used to predict the label, or category, of an observation. For example, we can predict whether a bank transaction is fraudulent or not. As there are two outcomes here - a fraudulent transaction, or non-fraudulent transaction, this is known as binary classification. Regression is used to predict continuous values. For example, a model can use features such as number of bedrooms, and the size of a property, to predict the target variable, price of the property.

7. Naming conventions

Note that what we call a feature throughout the course, others may call a predictor variable or independent variable. Also, what we call the target variable, others may call dependent variable or response variable.

8. Before you use supervised learning

There are some requirements to satisfy before performing supervised learning. Our data must not have missing values, must be in numeric format, and stored as pandas DataFrames or Series, or NumPy arrays. This requires some exploratory data analysis first to ensure data is in the correct format. Various pandas methods for descriptive statistics, along with appropriate data visualizations, are useful in this step.

9. scikit-learn syntax

scikit-learn follows the same syntax for all supervised learning models, which makes the workflow repeatable. Let's familiarize ourselves with the general scikit-learn workflow syntax, before we explore using real data later in the chapter. We import a Model, which is a type of algorithm for our supervised learning problem, from an sklearn module. For example, the k-Nearest Neighbors model uses distance between observations to predict labels or values. We create a variable named model, and instantiate the Model. A model is fit to the data, where it learns patterns about the features and the target variable. We fit the model to X, an array of our features, and y, an array of our target variable values. We then use the model's dot-predict method, passing six new observations, X_new. For example, if feeding features from six emails to a spam classification model, an array of six values is returned. A one indicates the model predicts that email is spam, and a zero represents a prediction of not spam.

10. Let's practice!

Now let's check our understanding of supervised learning and how to perform it in scikit-learn.

This exercise is part of the course

Supervised Learning with scikit-learn

IntermediateSkill Level

4.4+

Start Course for Free

In this chapter, you'll be introduced to classification problems and learn how to solve them using supervised learning techniques. You'll learn how to split data into training and test sets, fit a model, make predictions, and evaluate accuracy. You’ll discover the relationship between model complexity and performance, applying what you learn to a churn dataset, where you will classify the churn status of a telecom company's customers.

Exercise 1: Machine learning with scikit-learn

Current Exercise

Exercise 2: Binary classification Exercise 3: The supervised learning workflow Exercise 4: The classification challenge Exercise 5: k-Nearest Neighbors: Fit Exercise 6: k-Nearest Neighbors: Predict Exercise 7: Measuring model performance Exercise 8: Train/test split + computing accuracy Exercise 9: Overfitting and underfitting Exercise 10: Visualizing model complexity

In this chapter, you will be introduced to regression, and build models to predict sales values using a dataset on advertising expenditure. You will learn about the mechanics of linear regression and common performance metrics such as R-squared and root mean squared error. You will perform k-fold cross-validation, and apply regularization to regression models to reduce the risk of overfitting.

Exercise 1: Introduction to regression Exercise 2: Creating features Exercise 3: Building a linear regression model Exercise 4: Visualizing a linear regression model Exercise 5: The basics of linear regression Exercise 6: Fit and predict for regression Exercise 7: Regression performance Exercise 8: Cross-validation Exercise 9: Cross-validation for R-squared Exercise 10: Analyzing cross-validation metrics Exercise 11: Regularized regression Exercise 12: Regularized regression: Ridge Exercise 13: Lasso regression for feature importance

Having trained models, now you will learn how to evaluate them. In this chapter, you will be introduced to several metrics along with a visualization technique for analyzing classification model performance using scikit-learn. You will also learn how to optimize classification and regression models through the use of hyperparameter tuning.

Exercise 1: How good is your model?Exercise 2: Deciding on a primary metric Exercise 3: Assessing a diabetes prediction classifier Exercise 4: Logistic regression and the ROC curve Exercise 5: Building a logistic regression model Exercise 6: The ROC curve Exercise 7: ROC AUC Exercise 8: Hyperparameter tuning Exercise 9: Hyperparameter tuning with GridSearchCV Exercise 10: Hyperparameter tuning with RandomizedSearchCV

Learn how to impute missing values, convert categorical data to numeric values, scale data, evaluate multiple supervised learning models simultaneously, and build pipelines to streamline your workflow!

Exercise 1: Preprocessing data Exercise 2: Creating dummy variables Exercise 3: Regression with categorical features Exercise 4: Handling missing data Exercise 5: Dropping missing data Exercise 6: Pipeline for song genre prediction: I Exercise 7: Pipeline for song genre prediction: II Exercise 8: Centering and scaling Exercise 9: Centering and scaling for regression Exercise 10: Centering and scaling for classification Exercise 11: Evaluating multiple models Exercise 12: Visualizing regression model performance Exercise 13: Predicting on the test set Exercise 14: Visualizing classification model performance Exercise 15: Pipeline for predicting song popularity Exercise 16: Congratulations