IAI Actuarial Core Principles · Risk Modelling and Survival Analysis · Elementary principles of machine learning
A model fitted to training data gives a very low training error but a much higher error on a separate validation set. What is the most likely explanation and the usual remedy?
The model is overfitting. It fits noise in the training data, so it performs well there but poorly on unseen data. The usual remedies are simplifying the model, applying regularisation, or using more data, not adding parameters.
- AUnderfitting; add more flexibility to the model
- BOverfitting; simplify the model or regularise itCorrect
- CUnderfitting; remove the validation set
- DOverfitting; increase the number of parameters
- Bias in the validation set; discard it
Explanation
A large gap between low training error and high validation error indicates the model has captured noise specific to the training data, which is overfitting. Reducing complexity, regularisation or more data helps. Adding parameters would worsen it.
Did you get it right without looking?
One question tells you little. A timed set on Elementary principles of machine learning shows your real accuracy, how long you take and where you lose marks.
More Elementary principles of machine learning questions
- Which of the following is an example of a regression task, as opposed to a classification task, in a supervised learning setting?
- A general insurer in India wants to group its motor policyholders into segments using only rating variables such as vehicle age, engine capa…
- A general insurer wants to predict the claim amount (in rupees) for each motor policy from rating factors such as vehicle age, engine capaci…
- A lasso (L1-penalised) regression is fitted to predict lapse rates and the penalty parameter lambda is increased from a small value to a ver…
- Which statement about the K-means algorithm is correct?
- A lapse-prediction model is fitted to 1,000 policies. Of the 100 policies that actually lapsed, the model flags 70 as lapses. Of the 900 pol…