FRM Part I · FRM Exam Part I · Machine-Learning Methods
A risk manager wants to reduce a large set of highly correlated yield curve variables to a few uncorrelated factors before building a model. Which feature of principal components analysis makes it suited to this purpose?
PCA creates orthogonal, uncorrelated components ranked by the variance each explains, so a few components can summarize many correlated yield curve variables. It does not use a target variable, does not assign observations to groups, and works best when inputs are correlated.
- AIt produces orthogonal components ordered by the share of variance each explainsCorrect
- BIt maximizes the correlation between each component and a default indicator
- CIt assigns each observation to exactly one group
- DIt requires the original variables to be uncorrelated first
Explanation
PCA creates uncorrelated (orthogonal) linear combinations ranked by explained variance, so a few components can capture most variation in correlated inputs. It is unsupervised and uses no default indicator. Assigning observations to groups describes clustering, and PCA is most useful precisely when inputs are correlated.
Did you get it right without looking?
One question tells you little. A timed set on Machine-Learning Methods shows your real accuracy, how long you take and where you lose marks.
More Machine-Learning Methods questions
- A deep neural network for credit scoring achieves very low training error but much higher validation error. Which single action is most dire…
- A neuron has a sigmoid activation, f(z) = 1/(1+e^(-z)). Its inputs are x1 = 1 and x2 = 3, with weights w1 = 2 and w2 = -1, and bias b = 1. W…
- A ridge regression is fitted with a single standardized predictor and no intercept. The ordinary least squares slope is 0.80, the sum of squ…
- In k-means clustering, an analyst increases k from 3 to 4 on the same dataset and re-runs the algorithm to convergence. Which outcome is mos…
- A single weight w in a network is updated by gradient descent. The current weight is 0.80, the learning rate is 0.10, and the partial deriva…
- A data scientist uses 5-fold cross-validation on a dataset of 1,000 observations to select a model. Fold validation mean squared errors are …