Advanced Topics in Machine Learning Project

MSc in Artificial Intelligence

Aristotle University of Thessaloniki

1. Class Imbalance

a. SMOTE

b. Tomek Links

c. NearMiss

Class Imbalance Jupyter Notebook :

1. github : https://github.com/constseche/Advanced-ML/blob/master/Class_Imbalance/Class_Imbalance.ipynb
2. nbviewer : https://nbviewer.jupyter.org/github/constseche/Advanced-ML/blob/master/Class_Imbalance/Class_Imbalance.ipynb

2. Cost-Sensitive Learning

a. Class Weighting

b. Stratification aka Rebalancing

  a.	Under-Sampling 
  b.	Over – Sampling
  c.	Combination of the above 

c. Costing-Rejection Sampling combining with hard voting (Ensemble models)

d. Minimizing the cost: (this is used only in binary classification problem because some costcla features don’t offer possibilities for multiclass cost-matrix)

  a.	Without calibration
  b.	With sigmoid calibration 
  c.	With isotonic calibration

Cost-Sensitive Learning Jupyter Notebook :

A) Binary

1. github : https://github.com/constseche/Advanced-ML/blob/master/Cost_Sensitive_Learning/cost_sensitive_learning_binary.ipynb
2. nbviewer : https://nbviewer.jupyter.org/github/constseche/Advanced-ML/blob/master/Cost_Sensitive_Learning/cost_sensitive_learning_binary.ipynb

B) Multiclass

1. github : https://github.com/constseche/Advanced-ML/blob/master/Cost_Sensitive_Learning/cost_sensitive_learning.ipynb
2. nbviewer : https://nbviewer.jupyter.org/github/constseche/Advanced-ML/blob/master/Cost_Sensitive_Learning/cost_sensitive_learning.ipynb

3. Machine Learning Explainability

a. Permutation Importance

What features does a model think are important ? Which features might have a greater impact on the model predictions than the others ?

b. Partial Dependence Plots

How does each feature affect your predictions ?

c. SHAP Values

Understanding individual predictions.

Machine Learning Explainability Jupyter Notebook :

A) Heart Disease Dataset

1. github : https://github.com/constseche/Advanced-ML/blob/master/Interpretability/Heart%20Disease%20Machine%20Learning%20Interpretability.ipynb
2. nbviewer : https://nbviewer.jupyter.org/github/constseche/Advanced-ML/blob/master/Interpretability/Heart%20Disease%20Machine%20Learning%20Interpretability.ipynb

B) Fetal Health Dataset

1. github : https://github.com/constseche/Advanced-ML/blob/master/Interpretability/explainable_fetal.ipynb
2. nbviewer : https://nbviewer.jupyter.org/github/constseche/Advanced-ML/blob/master/Interpretability/explainable_fetal.ipynb

suggested in order to read big jupyter notebooks : https://nbviewer.jupyter.org/