Data Info | Xgboost model creation | Model visualization | Interactions | Variables' and interactions’ importance | Explanation of single prediction including interactions

Data Info | Xgboost model creation | Model visualization | Interactions | Variables' and interactions’ importance | Explanation of the single prediction including interactions

Overview | Example: Multiclass XGBoost model | Similar for LightGBM (only code) | Or for a random forest with | Example: SHAP subgroup analysis | Example: Different models | Fit linear regression and use {kernelshap} to get SHAP values

Overview | Installation | Usage | Compact SHAP analysis | Decompose single predictions | SHAP Interactions | Interface to other packages | LightGBM | fastshap | shapr | H2O | treeshap | DALEX | kernelshap | Any other package | References

Normal case | XGBoost | LightGBM | Probabilistic classification | LightGBM (binary probabilistic)

Setting | A first example | Two modifications | References

Introduction | Data and forest | Distribution of minimal depth | Various variable importance measures | Multi-way importance plot | Compare measures using ggpairs | Compare different rankings | Variable interactions | Conditional minimal depth | Prediction of the forest on a grid | Explain the forest

Model and explainer creation | Making predictions | Measuring performance | Global explanations | Variable importance | Partial dependence | Local explanations | Local variable attributions | SurvSHAP(t) | SurvLIME | Ceteris paribus

Introduction | Python library shap | Load data sets | Let's build models | Here shapper starts | Plotting results | Let's filter data for plot

Automatic explainer creation | Manual explainer creation | Helpful utility functions | References

Introduction | iBreakDown Package | Describing an explanation | Parameters of describe() function | Adjusting nonsignificance treshold | Adjusting label of the explanation | Short descriptions | Displaying values | Displaying numbers | Distribution details | SHAP | Combining all the parameters

Model, explanation & bias | Bias mitigation strategies | Pre-processing techniques | Distribution changing | Reweighting | Resampling | Post-processing techniques | ROC pivot | Cutoff manipulation | Tradeoff between bias and accuracy | Checking fairness on a test set | Summary | References

fairmodels | Why? | Data | Basic features | fairness check | plot density | Metric scores plot | fairness object - idea | What consists of fairness object? | Choosing best model | Stacked Barplot | Plot metric | Plot fairness PCA | Plot Heatmap | Metric and Performance Plot | Group Metric | Radar plot | Custom cutoff | All cutoffs | Ceteris paribus cutoff | Summary | References

Data | Models | Preparation for analysis of performance | Model ranking radar plot | Other methods

Data | Models | Preparation for residual (error) analysis | Plots | Observed vs predicted | Residuals vs observed, fitted or variable values | Density of residuals | Boxplot of residuals | Autocorrelation function of residuals | Autocorrelation of residuals | Correlation of models | Principal component analysis (PCA) of models | Regression error characteristic curve (REC) | Regression receiver operating characteristic (RROC) | Scale location | Two-sided empirical cumulative distribution function (TSECDF) | Other methods

Data | Models | Preparation for analysis of observation influence | Plot of Cook's distances | Other methods

Data | Models | Preparation for evaluation analysis | Plots | Receiver operating characteristic (ROC) | LIFT chart | Other methods

Use-case - regression problem | Half-normal plot | Binomial model | Use-cases (classification) | Other methods | References

R & Python Examples | R | mlr dashboard | mlr3 dashboard | xgboost dashboard | caret dashboard | h2o dashboard | parsnip dashboard | tidymodels dashboard | Python | scikit-learn dashboard | lightgbm dashboard | keras/tensorflow dashboard | References

modelStudio parameters | instance explanations | grid size | animations | more calculations means more time | no EDA mode | progress bar | viewer or browser? | parallel computation | additional options | update observations | Shiny | DALEXtra | References

Introduction | Install shaper and shap | R package shapper | Python library shap | Load data sets | Let's build a model | Prediction to be explained | Here shapper starts | Plotting results

Building a black-box model | Creating an explainer | Creating a safe_extractor | Transforming data | Creating white-box models on original and transformed datasets | Comparing models performance

Building a black-box model | Creating an explainer | Creating a safe_extractor | Transforming data | Creating white-box models on original and transformed datasets | Comparing models performance

Introduction | ingredients Package | Feature Importance | Ceteris Paribus Profiles | Partial Dependence Profiles

Data for Titanic survival | Model for Titanic survival | Explainer for Titanic survival | Model Level Feature Importance | Feature effects | age | Partial Dependence Profiles | Conditional Dependence Profiles | Accumulated Local Effect Profiles | Instance level explanations | Profile clustering | Session info

Preface | Load data & libraries | Models | Create Explainers & Arena | Make explainers manualy

Data for Titanic survival | Model for Titanic survival | Explainer for Titanic survival | Break Down plot with D3 | Calculate variable attributions | Plot attributions with ggplot2 | Plot attributions with D3 | Calculate uncertainty for variable attributions | Show only top features | Force OX axis to be from 0 to 1

1 Dataset | 2 Random forest regression model | 3 Calculate Partial Dependence Profiles | 4 Calculate measure of global variable importance | 5 Comparison of the importance of variables for two or more models

1 Dataset | 2 Random forest regression model | 3 New observation | 4 Calculate Ceteris Paribus profiles | 5 Calculate measure of local variable importance | 6 Comparison of two or more methods of calculating the importance of variables | 7 Comparison of the importance of variables for two or more models

Setup | Basic use of Arena - Single Model, Global Explanations | Train a model | Create an explainer | Create an Arena and add the model | Run the live Arena server | Intermediate use of Arena - Single Model, Global and Local Explanations | Add local explanations | Advanced use of Arena - Multiple Models, Global and Local Explanations | Create more models | Create explainers | Add more models to the Arena | Serverless version of the Arena

Setup | Prepare models | Run arena | Custom observation names | Appending data

Setup | Prepare models | Prepare observations | Create arena | Appending data | New Arena session | Append to already existing session

Simulated data | Explainer for the models | Ceteris paribus | Dependence profiles | Dependence profiles in groups | Session info

Sampling for local exploration | Fitting black box model | Fitting the explanation model