Assumption check

As explained in Machine Learning System Design

An assumption check is a procedure for assessing whether the conditions required by a statistical or machine learning model are supported by data or diagnostics.

An assumption check tests whether a model’s expected prediction, bias, and residual behavior is supported by data. Unexpected patterns can motivate different models, more data, or better features. section 9.2.2

For linear regression, check these assumptions:

• Linearity between predictors and the target. section 9.2.2
• Strict exogeneity, expressed as zero-centered residuals. section 9.2.2
• Normally distributed residuals. section 9.2.2
• Constant residual variance across predictor values. section 9.2.2
• Independent residual error terms and no highly influential observations. section 9.2.2

Demand forecasting

After fitting a linear regression for demand forecasting, inspect residuals with diagnostic plots. Patterns, biased residuals, changing variance, or influential points indicate that the model or its inputs need reconsideration. section 9.2.2

Figure 9.10 Four diagnostic plots for linear regression’s residual analysis in both cases (Case 1: Assumptions are met)

Compare four residual diagnostic plots for a case where the linear-regression assumptions are met. section 9.2.2

How Machine Learning System Design covers it

How deep it goes

Defines five linear-regression assumptions and checks them with residual diagnostic plots, while deferring broader pipeline testing.

Examples

Uses demand forecasting with linear regression and paired residual diagnostic plots showing met and unmet assumptions.

What you can do after

Evaluate linear-regression assumptions through residual plots and encode expected prediction or residual behavior as unit tests.

How it connects

Connects model selection, data collection, feature engineering, residuals, unit tests, linear regression, and diagnostic plots.

liveBookmanning.com

Read Machine Learning System Design on liveBook Get the book at manning.com

Related concepts

• check for assumptionscontrasts with · A prompting technique that instructs a large language model to recognize when required conditions are unmet and provide an explicit fallback instead of guessing.

This explainer was generated by Manning's knowledge graph from the text of Machine Learning System Design. The section links point to the parts of the book it draws on.