/>
X

Pitfalls to Avoid when Interpreting Machine Learning Models

Modern requirements for machine learning models include both high predictive performance and model interpretability. A team of experts in explainable AI highlights pitfalls to avoid when addressing model interpretation, and discusses open issues for further research. Images created by Cristoph Molnar: https://twitter.com/ChristophMolnar/status/1281272026192326656
1-bad-model-generalization.jpg
1 of 8 Christoph Molnar

1-bad-model-generalization

Any interpretation of relationships in the data is only as good as the model it is based on. Both under- and overfitting can lead to bad models with a misleading interpretation.

=> Use proper resampling techniques to assess model performance.

2-unnecessary-use-of-ml.jpg
2 of 8 Christoph Molnar

2-unnecessary-use-of-ml

Don't use a complex ML model when a simple model has the same (or better) performance or when the gain in performance would be irrelevant. 

=> Check the performance of simple models first, gradually increase complexity.

3-1-ignoring-feature-dependence.jpg
3 of 8 Christoph Molnar

3-1-ignoring-feature-dependence

When features depend on each other (as they usually do) interpretation becomes tricky, since effects can't be separated easily. 

=> Analyze feature dependence. Be careful with the interpretation of dependent features. Use appropriate methods.

3-2-confusing-dependence-with-correlation.jpg
4 of 8 Christoph Molnar

3-2-confusing-dependence-with-correlation

Correlation is a special case of dependence. The data can be dependent in much more complex ways.

=> In addition to correlation, analyze data with alternative association measures such as HSIC.

4-misleading-effects-due-to-interaction.jpg
5 of 8 Christoph Molnar

4-misleading-effects-due-to-interaction

Interactions between features can "mask" feature effects. 

=> Analyze interactions with e.g. 2D-PDP and the interactions measures.

5-ignoring-estimation-uncertainty.jpg
6 of 8 Christoph Molnar

5-ignoring-estimation-uncertainty

There are many sources of uncertainty: model bias, model variance, estimation variance of the interpretation method. 

=> In addition to point estimates of (e.g., feature importance) quantify the variance. Be aware of what is treated as 'fixed.'

6-ignoring-multiple-comparisons.jpg
7 of 8 Christoph Molnar

6-ignoring-multiple-comparisons

If you have many features and don't adjust for multiple comparisons, many features will be falsely discovered as relevant for your model. 

=> Use p-value correction methods.

7-unjustified-causal-interpretation.jpg
8 of 8 Christoph Molnar

7-unjustified-causal-interpretation

Per default, the relationship modeled by your ML model may not be interpreted as causal effects.

=> Check whether assumption can be made for a causal interpretation.

Related Galleries

Remote working, data breaches, job vacancies, 5G devices, and more: ZDNet's research roundup
istock-1210523597.jpg

Related Galleries

Remote working, data breaches, job vacancies, 5G devices, and more: ZDNet's research roundup

11 Photos
Nvidia Jetson Nano 2GB Development Kit With Wi-Fi Bundle
Nvidia Jetson Nano 2GB Development Kit

Related Galleries

Nvidia Jetson Nano 2GB Development Kit With Wi-Fi Bundle

19 Photos
Azure Synapse Analytics data lake features: up close
azure-ml-integration-via-notebook.png

Related Galleries

Azure Synapse Analytics data lake features: up close

19 Photos
Over-hyped AI, data security worries and an e-commerce boom: Tech research round-up
Young Man looking at digital tablet

Related Galleries

Over-hyped AI, data security worries and an e-commerce boom: Tech research round-up

10 Photos
Pitfalls to Avoid when Interpreting Machine Learning Models
1-bad-model-generalization.jpg

Related Galleries

Pitfalls to Avoid when Interpreting Machine Learning Models

8 Photos
AAEON BOXER-8120AI
AAEON BOXER-8120AI

Related Galleries

AAEON BOXER-8120AI

6 Photos
IT spending, cloud computing, big data, virtual reality, and more: Research round-up
racial-and-ethnic-composition-by-religious-group.png

Related Galleries

IT spending, cloud computing, big data, virtual reality, and more: Research round-up

12 Photos