This repository contains references to the Medium articles, available in https://medium.com/@vinicius_trevisan
Note: some articles may be in Brazilian Portuguese.
Article: Comparing robustness of MAE MSE and RMSE [EN]
Explains how MAE, MSE and RMSE metrics for regression evaluation are affected by the presence of outliers, then performs an experiment to illustrate this effect.
__
Article 1: Interpreting ROC Curve and ROC AUC for Classification Evaluation [EN]
What are ROC Curves, how to interpret and use them to evaluate binary classification models. How to build ROC Curves from scratch. What is Area Under ROC Curve (ROC AUC / AUROC) and how it is used to evaluate binary classification models.
Article 2: Multiclass classification evaluation with ROC Curves and ROC AUC [EN]
How to use the "One versus Rest" (OvR) and "One versus One" (OvO) strategies to adapt ROC Curves and ROC AUC for multiclass classification.
__
Article 1: Using SHAP Values to explain how your machine learning model works [EN]
What are SHAP values and how they explain the effect each feature of the model has on the prediction. How to use the SHAP library and their plots. How SHAP values are calculated.
Article 2: Boruta SHAP: an amazing tool for feature selection every data scientist should know [EN]
What is Boruta and why is it a robust feature selection method. How to make Boruta better by using SHAP as a feature importance score. Using the Boruta-SHAP library to perform the feature selection.
__
Article 1: Comparing sample distributions with Kolmogorov-Smirnov test [EN]
Introduction to the Kolmogorov-Smirnov (KS) test, and how it is used in normality tests. How to use KS tests to compare two samples and discover if they come from the same distribution.
Article 2: Evaluating classification models with Kolmogorov-Smirnov test [EN]
How to use the KS test as a tool to evaluate class separability on binary classifiers. Experimental approach comparing three different classification scenarios.
__
Article: Evaluating the potential return of a model with Lift, Gain, and Decile Analysis [EN]
This article shows how to use Lift, Gain, and Decile Analysis to understand the benefit and return of our classification models.
__
Article: Checking model stability and population shift with PSI and CSI [EN]
Shows how to use PSI (Population Stability Index) and CSI (Characteristic Stability Index) to verify the stability and data shift of the model in production.
__
Article: Dealing with Categorical Variables by using Target Encoder [EN]
Introduction to the idea behind Target Encoder, code examples and potential problems when using it. How to use Target Encoding in multiclass scenarios.
__
Article:
Understanding Convolutional Neural Networks (CNNs) [EN] (TBD)
Como funcionam as Redes Neurais Convolucionais (CNNs) [PT-BR]
Fundamentals of CNNs, and a quick explanation of how they work and how features such as kernel size, stride and padding affect their output.
__