sameervk/UCI_Breast_Cancer_Classification

ML of UCI Breast Cancer Dataset

UCI Breast Cancer Classification using scikit-learn and XGBoost

*  Exploratory data analysis using pandas, seaborn plotting and 
   scikit-learn feature selection
*  Data preprocessing including splitting, scaling and balancing 
   data (SMOTEENN, SVMSMOTE algorithms)
*  Classification using Logistic Regression, Random Forest and 
   XGBClassifier
*  Hyperparameter tuning of XGBClassifier parameters using 
   RandomizedSearchCV of scikit-learn
*  Plotting of the train, test and fit data
-----
*  The best training accuracy (~99.8%) comes from a XGBClassifier on 
   SVMSMOTE balanced data but with a test accuracy of only 90.6% indicating significant overfitting. 
*  The best test accuracy (~95.0%) comes from Logistic Regression, 
   RandomForestClassifier and XGBClassifier all fit on SMOTEENN balanced data. 
*  Possible suggestions for improvement: need further exploratory data analysis to examine for outliers 
   and leverage points which can be done using studentized residuals.