Irrelevant feature

Q: How deeply does Machine Learning for Tabular Data: XGBoost, Deep Learning, and AI cover irrelevant feature?

Defines irrelevance, gives domain and univariate checks, and explains reshuffling-based assessment; detailed feature-selection tools are deferred.

As explained in Machine Learning for Tabular Data: XGBoost, Deep Learning, and AI

An irrelevant feature is an input attribute with little or no association with the target and little useful information for the prediction task.

An irrelevant feature is an input attribute with little or no association with the target and little useful information for the prediction task. In tabular data, it adds noise rather than reliable signal. section 2.2.3

Assess candidate features before retaining them:

Use domain knowledge to rule out features that are implausible sources of information. section 2.2.3
Apply simple univariate tests, such as correlation or chi-square analysis, as an initial statistical check. section 2.2.3
Build a model, reshuffle each feature, and compare performance; features whose randomization changes little are likely irrelevant or redundant. section 2.2.3
Drop features that contribute no useful signal to reduce false disturbances, memory use, and training and prediction time. section 2.2.3

Randomization check

Train a model on the available features, randomly replace the values of one feature, and observe the change in predictive performance. Little change suggests that the feature can be removed, although the same result may indicate redundant information elsewhere. section 2.2.3

How Machine Learning for Tabular Data: XGBoost, Deep Learning, and AI covers it

Cover of Machine Learning for Tabular Data: XGBoost, Deep Learning, and AI

liveBook manning.com

Read Machine Learning for Tabular Data: XGBoost, Deep Learning, and AI on liveBook Get the book at manning.com

Related concepts

attributekind of · An attribute is a named property, value, or piece of information associated with an entity, object, or program element.
overfittingproduces · Overfitting is a modeling condition in which a model fits patterns or idiosyncrasies in its training data too closely and consequently generalizes poorly to new data.
feature selectionaddressed by · Feature selection is the process of choosing a subset of input features for use in a modeling, analysis, or prediction task.
tree ensemblecontrasts with · A tree ensemble is a predictive model that combines multiple decision trees or tree-derived models to produce predictions.

This explainer was generated by Manning's knowledge graph from the text of Machine Learning for Tabular Data: XGBoost, Deep Learning, and AI. The section links point to the parts of the book it draws on.