Background:

there are many computational tools for managing images and data sets; reducing the size of these favors the management of information.

Objective:

reduce the data set size for better information management.

Methods:

the Breast Cancer Wisconsin data set (biopsy information - nuclear cells) and the Python Jupyter platform were used. Principal Component Analysis (PCA) and Kernel PCA (kPCA) techniques were implemented to reduce the dimension to 2, 4, 6. Cross-validation was made to select the best hyperparameters of the regression and support vector machine algorithms Logistics. The classification was carried out with the original training test, training test (PCA and kPCA) and training test (data transformed from PCA and kPCA). Accuracy, precision, completeness, recovery, and area under the curve were analyzed.

Results:

the PCA with six components explained the variation rate by almost 90%. The best hyperparameters found for the vector support machine: linear kernel and C = 100, for logistic regression were C = 100, Newton-cg solution (solver) and I2. The best results of the metrics were for PCA 2 and 4 (0.99, 0.99, 1, 0.99, 0.99). For the training set with original data they were 0.96; 0.95; 0.99; 0.97; 0.95. For logistic regression the best results were for kPCA with 6 components. The statistical results were equal to 1. For the training set with original data, these values were 0.96; 0.95; 0.99; 0.97; 0.95.

Conclusions:

the results of the metrics improved using PCA and kPCA.

        · abstract in Spanish     · text in Spanish     · Spanish ( pdf )