Linear Discriminant Analysis (deprecated)

You can use Fisher Linear Discriminant Analysis to create a set of scores that identifies the combination of features that best separate two or more classes.

Warning
This module is provided for backward compatibility with experiments created using the pre-release version of Azure Machine Learning, and will soon be deprecated. We recommend that you modify your experiments to use Fisher Linear Discriminant Analysis instead.

Linear discriminant analysis is often used for dimensionality reduction, because it projects a set of features onto a smaller feature space while preserving the information that discriminates between classes. This not only reduces computational costs for a given classification task, but can help prevent overfitting.

You provide a set of possible feature columns as inputs, and the algorithm determines the optimal combination of the input columns that linearly separates each group of data while minimizing the distances within each group.

Linear discriminant analysis is similar to analysis of variance (ANOVA) in that it works by comparing the means of the variables, and is based on these assumptions:

• Predictors are independent

• Values are normally distributed

• Variances among groups are similar

Note that Linear Discriminant Analysis is sometimes abbreviated to LDA, but this is easily confused with Latent Dirichlet Allocation. The techniques are completely different, so in this documentation, the acronym LDA will be used only for Latent Dirichlet Allocation.

To see examples of how feature selection is used in machine learning experiments, see these sample experiments in the Model Gallery:

• The Twitter Sentiment Analysis sample uses filter-Based Feature Selection to improve experiment results.

• The Fisher Linear Discriminant Analysis sample demonstrates how to use this module for dimensionality reduction.

• This method works only on continuous variables, not categorical or ordinal variables.

• Rows with missing values are ignored when computing the transformation matrix.

• The algorithm will examine all numeric columns not designated as labels, to see if there is any correlation. If you want to exclude a numeric column, add a Project Columns module before feature selection to create a view that contains only the columns you wish to analyze.