第三个方法不理解

问题如下：

The most appropriate method to resolve the computation problem discussed by Lovell and his colleague is:

选项：

A.Method 1.

B.Method 2.

C.Method 3.

解释：

A is correct. Method 1 is the most appropriate method to resolve the computation problem. It uses principal components analysis. Principal components analysis is a common methodology of complex or large dataset models with a significant number of features to be reduced to their common behavioral attributes within a model that have the strongest predictive power of the defined outcome (i.e., consumer default). Since the model reduces the number of variables needed to explain the variation in the data, it reduces the computational time required to complete it instead of using every single parameter for every single record.

B is incorrect. Method 2 suggests decreasing the learning rate of the algorithm. Decreasing the learning rate would actually increase the computational needs because it would increase the number of iterations that the model would need to run to be able to “learn” the stated objective.

C is incorrect. Method 3 applies winsorization of the existing data. Winsorization is a data wrangling technique used to manage outlier scenarios by replacing the individual outlier with the minimum or maximum of non-outlier data points—effectively adding to the ends of a distribution curve. This methodology removes variations, but the data points still remain, so it would have no effect on the computational needs of the model.