IAI Actuarial Core Principles · Risk Modelling and Survival Analysis · Elementary principles of machine learning
An analyst uses 5-fold cross-validation on 1,000 policy records to choose between candidate models. Which description of the procedure is correct?
In 5-fold cross-validation the data are divided into five parts, and each part is used once as the validation set while the model is trained on the other four. The five validation errors are then averaged to estimate out-of-sample performance.
- AThe data are split into 5 parts; each part is used once as the validation set while the model is trained on the other 4, and the 5 errors are averagedCorrect
- BThe data are split into 5 parts; the model is trained on one part and tested on the other 4, and this is done once
- CThe model is trained 5 times on all 1,000 records using different random starting values, and the errors are averaged
- DThe data are split into 5 parts and the model is trained and tested on each part separately, with the largest error reported
- The 1,000 records are resampled with replacement 5 times and the model is tested on the records that were drawn
Explanation
In k-fold cross-validation each fold serves once as validation data while the remaining k-1 folds train the model. The k validation errors are then averaged to estimate out-of-sample performance. Training on one fold and testing on four reverses the usual roles and is not k-fold.
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
- 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 likel…
- A data scientist fits a very flexible model to a claims dataset. It achieves a very low error on the training data but a much higher error o…
- A analyst applies K-means clustering to policyholder data containing annual premium (in rupees, values in the lakhs) and age (in years, valu…
- A claims-classification model is trained with early stopping. Training error keeps decreasing each iteration, while validation error decreas…
- A health insurer in Pune has records of 50,000 past policyholders, each with age, BMI, smoking status and a recorded label showing whether a…
- In 5-fold cross-validation on 500 observations used to compare two regularised regression models, which statement is correct?