AIExplainer
Machine Learning Intermediate 2 min read

What is oversampling?

Collecting more data than needed to improve model performance and reduce bias

Oversampling is a technique used in machine learning to collect more data than necessary to improve the performance and accuracy of a model. This is often done to reduce bias in the data and ensure that the model is trained on a diverse range of examples.

Oversampling is like taking extra photos of a scene to ensure you capture the perfect shot. Just as having more photos gives you a better chance of getting the one you want, oversampling gives a model a better chance of learning from the data and making accurate predictions.

A company building a facial recognition model might collect a large dataset of images of people from different ethnic backgrounds to ensure that the model is not biased towards any particular group.

Oversampling is used in machine learning to improve the performance of models, particularly in cases where the data is imbalanced or biased. It is often used in conjunction with other techniques, such as undersampling and data augmentation.

One common misconception about oversampling is that it always leads to better model performance. However, oversampling can also lead to overfitting, where the model becomes too specialized to the training data and fails to generalize well to new data.

Oversampling has been used in machine learning for many years, but its importance has grown in recent years as the field has become more focused on issues of bias and fairness.

data augmentation resampling bootstrap sampling

Three products for different needs — explore what’s relevant to you.