rarefun: Functions for Rare Events Analysis

Overview

rarefun is an R package that provides a collection of functions for detecting and analyzing rare events in datasets. The package includes implementations of various anomaly detection methods and tools for spatial and temporal analysis of rare events.

Features

• Anomaly Detection Methods:

  • Removing trend (and seasonality) to analyze rare events in residuals (rare_residuals), including the following techniques:
  • Isolation forest (rare_iforest)
  • DBSCAN clustering (rare_dbscan)

• Spatial Analysis:

  • Spatial point matching based on geographic corrdinates (match_spatial_points), including fast nearest neighbor search

• Statistical Methods:

  • Matthews Correlation Coefficient (mcc), including bootstrap confidence intervals
  • Goodness-of-fit metrics (R1, ATWE) for quantile regression (gof_qr)
  • Partial dependence plots for quantile random forest (partial_qrf)
  • Kneedle algorithm for knee point detection (kneedle)

Installation

You can install the development version of rarefun from GitHub:

# install.packages("devtools")
# devtools::install_github("vlyubchich/rarefun")
install.packages("rarefun", repos = c('https://vlyubchich.r-universe.dev', 'https://cloud.r-project.org'))

Usage

library(rarefun)

# Example: Detect anomalies using Isolation Forest
set.seed(123)
data <- matrix(rnorm(1000), nrow = 500)
data[1:5, ] <- data[1:5, ] + 10  # Add some outliers

result <- rare_iforest(data, ntrees = 100, threshold = 0.7)
table(result$is_anomaly)

Authors

• Vyacheslav Lyubchich (maintainer) - ORCID: 0000-0001-7936-4285
• Geneviève Nesslage - ORCID: 0000-0003-1770-6803

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Bug reports

If you encounter any bugs or issues, please report them at: https://github.com/vlyubchich/rarefun/issues