Concept Drift and Concept Shift Detection for Predictive Models

Concept drift refers to the change in the data distribution or in the relationships between variables over time. 'drifter' calculates distances between variable distributions or variable relations and identifies both types of drift. Key functions are: calculate_covariate_drift() checks distance between corresponding variables in two datasets, calculate_residuals_drift() checks distance between residual distributions for two models, calculate_model_drift() checks distance between partial dependency profiles for two models, check_drift() executes all checks against drift. 'drifter' is a part of the 'DrWhy.AI' universe (Biecek 2018) .

Datasets and Functions Used in the Book 'Przewodnik po Pakiecie R'

Data sets and functions used in the polish book "Przewodnik po pakiecie R" (The Hitchhiker's Guide to the R). See more at < http://biecek.pl/R>. Among others you will find here data about housing prices, cancer patients, running times and many others.

Gaussian Mixture Modeling Algorithms and the Belief-Based Mixture Modeling

Two partially supervised mixture modeling methods: soft-label and belief-based modeling are implemented. For completeness, we equipped the package also with the functionality of unsupervised, semi- and fully supervised mixture modeling. The package can be applied also to selection of the best-fitting from a set of models with different component numbers or constraints on their structures. For detailed introduction see: Przemyslaw Biecek, Ewa Szczurek, Martin Vingron, Jerzy Tiuryn (2012), The R Package bgmm: Mixture Modeling with Uncertain Knowledge, Journal of Statistical Software .

Model Agnostic Explainers for Individual Predictions

Model agnostic tool for decomposition of predictions from black boxes. Break Down Table shows contributions of every variable to a final prediction. Break Down Plot presents variable contributions in a concise graphical way. This package work for binary classifiers and general regression models.

Ceteris Paribus Profiles

Ceteris Paribus Profiles (What-If Plots) are designed to present model responses around selected points in a feature space. For example around a single prediction for an interesting observation. Plots are designed to work in a model-agnostic fashion, they are working for any predictive Machine Learning model and allow for model comparisons. Ceteris Paribus Plots supplement the Break Down Plots from 'breakDown' package.

Tools for Storing, Restoring and Searching for R Objects

Data exploration and modelling is a process in which a lot of data artifacts are produced. Artifacts like: subsets, data aggregates, plots, statistical models, different versions of data sets and different versions of results. The more projects we work with the more artifacts are produced and the harder it is to manage these artifacts. Archivist helps to store and manage artifacts created in R. Archivist allows you to store selected artifacts as a binary files together with their metadata and relations. Archivist allows to share artifacts with others, either through shared folder or github. Archivist allows to look for already created artifacts by using it's class, name, date of the creation or other properties. Makes it easy to restore such artifacts. Archivist allows to check if new artifact is the exact copy that was produced some time ago. That might be useful either for testing or caching.

The Proton Game

'The Proton Game' is a console-based data-crunching game for younger and older data scientists. Act as a data-hacker and find Slawomir Pietraszko's credentials to the Proton server. You have to solve four data-based puzzles to find the login and password. There are many ways to solve these puzzles. You may use loops, data filtering, ordering, aggregation or other tools. Only basics knowledge of R is required to play the game, yet the more functions you know, the more approaches you can try. The knowledge of dplyr is not required but may be very helpful. This game is linked with the ,,Pietraszko's Cave'' story available at http://biecek.pl/BetaBit/Warsaw. It's a part of Beta and Bit series. You will find more about the Beta and Bit series at http://biecek.pl/BetaBit.

Mini Games from Adventures of Beta and Bit

Three games: proton, frequon and regression. Each one is a console-based data-crunching game for younger and older data scientists. Act as a data-hacker and find Slawomir Pietraszko's credentials to the Proton server. In proton you have to solve four data-based puzzles to find the login and password. There are many ways to solve these puzzles. You may use loops, data filtering, ordering, aggregation or other tools. Only basics knowledge of R is required to play the game, yet the more functions you know, the more approaches you can try. In frequon you will help to perform statistical cryptanalytic attack on a corpus of ciphered messages. This time seven sub-tasks are pushing the bar much higher. Do you accept the challenge? In regression you will test your modeling skills in a series of eight sub-tasks. Try only if ANOVA is your close friend. It's a part of Beta and Bit project. You will find more about the Beta and Bit project at < https://github.com/BetaAndBit/Charts>.

LIME-Based Explanations with Interpretable Inputs Based on Ceteris Paribus Profiles

Local explanations of machine learning models describe, how features contributed to a single prediction. This package implements an explanation method based on LIME (Local Interpretable Model-agnostic Explanations, see Tulio Ribeiro, Singh, Guestrin (2016) ) in which interpretable inputs are created based on local rather than global behaviour of each original feature.

Gaussian Model Invariant by Permutation Symmetry

Find the permutation symmetry group such that the covariance matrix of the given data is approximately invariant under it. Discovering such a permutation decreases the number of observations needed to fit a Gaussian model, which is of great use when it is smaller than the number of variables. Even if that is not the case, the covariance matrix found with 'gips' approximates the actual covariance with less statistical error. The methods implemented in this package are described in Graczyk et al. (2022) .