Summary of Imputation For Prediction: Beware Of Diminishing Returns, by Marine Le Morvan (soda) et al.
Imputation for prediction: beware of diminishing returns
by Marine Le Morvan, Gaël Varoquaux
First submitted to arxiv on: 29 Jul 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: Machine Learning (cs.LG); Machine Learning (stat.ML)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary This study investigates the relationship between imputation and predictive accuracy in machine learning models. The authors aim to determine whether advanced imputation methods yield significantly better predictions than simple constant imputation. They analyze 19 datasets, combining different imputation and predictive models, and find that imputation accuracy is less important when using expressive models or incorporating missingness indicators as inputs. However, it matters more for generated linear outcomes than real-data outcomes. Interestingly, the study shows that including a missingness indicator improves prediction performance even in cases where data is Missing Completely At Random (MCAR). Overall, the authors conclude that investing in better imputations may not significantly improve prediction performance on real-data with powerful models. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This research looks at how filling in missing values affects the accuracy of predictions made by machine learning models. The study wants to know if using more advanced methods for filling in missing values leads to better predictions. They tested different combinations of imputation and predictive models on 19 datasets and found that using simpler methods can be just as good as more complex ones, especially when using strong prediction models. However, they also found that including information about which values are missing can actually make the predictions more accurate. Overall, this study suggests that trying to fill in missing values better may not necessarily lead to much improvement in real-world predictions. |
Keywords
» Artificial intelligence » Machine learning
