- Load Libraries
- Load Dataset
- Data Exploration: Descriptive Statistics
- describe(): statistics - mean, stdev, min/max
- class distribution
- correlations
- skew
- data types
- Data Exploration: Visualization
- histogram
- density plot
- box whisker plot
- correlation matrix plot
- scatter plot matrix
- Data Cleaning
- Data Pre-Processing
- Rescale
- Standardize
- Normalize
- Binarize
- Feature Selection
- Univariate selection
- Recursive Feature Elimination
- Principal Component Analysis
- Feature Importance
- Data Transform
- Split-out validation dataset: Resampling
- Train-Test Data split
- k-fold cross validation
- LOOCV
- Repeated Random train-test splits
- Algorithm Performance Metrics
- Classification Metrics
- Classification Accuracy
- Logarithmic Loss
- Area under ROC curve
- Confusion Matrix
- Classification Report: precision, recall, F1 score, support
- Regression Metrics
- Mean Absolute Error
- Mean Squared Error
- R^2
- Classification Metrics
- Algorithm Performance
- Classification
- Linear Algorithms
- Logistic Regression
- Linear Discriminant Analysis
- Non-Linear Algorithms
- kNN
- Naive Bayes
- CART
- SVM (SVC)
- Linear Algorithms
- Regression
- Linear Algorithms
- Linear Regression
- Ridge Regression
- LASSO Linear Regression
- ElasticNet Regression
- Non-Linear Algorithms
- kNN
- CART
- SVM (SVR)
- Linear Algorithms
- Classification
- Algorithm Comparison
- Test harness to compare multiple algorithms efficiently on a single dataset
- Comparison visualization
- Preventing Data Leakage using Pipelines
- Data Preparation and Modeling Pipeling
- Feature Extractions and Modeling Pipeling
- Algorithm Tuning: Hyperparameter Optimization
- Grid Search Parameter Tuning
- Random Search Parameter Tuning
- Ensembles
- Bagging
- Bagged Decision Trees: Best for high variance algorithms
- Random Forests
- Extra Trees
- Boosting
- AdaBoost
- XGBoost: Stochastic Gradient Boosting (Gradient Boosting Machines - GBM)
- Voting
- Bagging
- Predictions on validation dataset
- Create standalone model on entire dataset
- Model persistence
- pickle
- joblib