Binomial logistic regression

Diagnostic breast cancer Wisconsin

By: William-CCS96

Introduction

This project aims to generate a binomial logistic regression model to classify, based on an examination, whether cancer is benign or malignant.

The work is carried out based on analysis, compression, data cleaning, metrics, testing and validation of the model, with the following work path:

• Development
• Understanding the data
• Data cleaning
  • Check null values
• Correlation analysis
• Data standardization:
• Exploratory data analysis
• Model creation
  • Split training and test data
  • Create and train the model
  • Evaluate the accuracy of the model with metrics
  • Evaluate the model with cross validation
• Analysis of results
  • Get the probabilities
  • Get the coefficients
  • Confusion Matrix
• Test
  • Test eliminating the characteristics with lower coefficients
  • Test balancing the number of class records with SMOTE
• Regularizers
  • L1 Lasso
  • L2 Ridge
• Conclusions