Detect Density-Based Anomalies (LOF)

Description

Uses the Local Outlier Factor (LOF) algorithm to identify anomalies based on local density. It is useful for detecting outliers in multi-dimensional data that Z-score misses.

Usage

detect_density(data, k = 5, threshold = 1.5)

Arguments

Value

A data frame with the outliers and their LOF score.

Examples

df <- data.frame(x = c(rnorm(50), 5), y = c(rnorm(50), 5))
detect_density(df, k = 5)

Detect Multivariate Anomalies (Mahalanobis Distance)

Description

Identifies outliers based on the relationship between multiple variables using Mahalanobis Distance. This is useful when individual values are normal, but their combination is anomalous (e.g., high weight for low height).

Usage

detect_multivariate(data, columns, confidence_level = 0.99)

Arguments

Value

A data frame with the multivariate outliers and their Mahalanobis distance.

Examples

# Generate dataset (n=50) with strong correlation
df <- data.frame(x = rnorm(50), y = rnorm(50))
df$y <- df$x * 2 + rnorm(50, sd = 0.5) # y depends on x

# Add an anomaly: normal x, but impossible y
anomaly <- data.frame(x = 0, y = 10)
df <- rbind(df, anomaly)

# Detect
detect_multivariate(df, columns = c("x", "y"))

Detect Anomalies in a Data Frame

Description

This function identifies rows containing outliers in a specific numeric column. It supports two methods:

• zscore: Based on the standard deviation (statistical approach). Best for normal distributions.

• iqr: Based on the Interquartile Range (robust approach). Best for data with skewness or extreme outliers.

Usage

detect_outliers(data, column, method = "zscore", threshold = 3)

Arguments

Value

A data frame containing only the rows considered outliers, with an additional column displaying the calculated score or bounds.

Examples

# Example with a clear outlier
df <- data.frame(
  id = 1:6,
  value = c(10, 12, 11, 10, 500, 11)
)

# Detect using IQR (Robust)
detect_outliers(df, column = "value", method = "iqr")

# Detect using Z-Score
detect_outliers(df, column = "value", method = "zscore")

Plot Outliers with ggplot2

Description

Visualizes the distribution of a variable and highlights detected outliers in red. It combines a boxplot (for context) and jittered points (for individual data visibility).

Usage

plot_outliers(data, column, method = "zscore", threshold = 3)

Arguments

Value

A ggplot object. You can add more layers to it using +.

Examples

library(ggplot2)
df <- data.frame(val = c(rnorm(50), 10)) # 50 normal points and one outlier
plot_outliers(df, "val", method = "iqr")

Scan Entire Dataset for Outliers

Description

Iterates through all numeric columns in the dataset and provides a summary table of outliers found.

Usage

scan_data(data, method = "iqr")

Arguments

Value

A summary data frame with columns: Column, Outlier_Count, and Percentage.

Examples

df <- data.frame(
  a = c(1:10, 100),
  b = c(1:10, 1)
)
scan_data(df, method = "iqr")

Treat Outliers (Winsorization/Capping)

Description

Instead of removing outliers, this function replaces extreme values with the calculated upper and lower boundaries (caps). This technique is often called "Winsorization".

Usage

treat_outliers(data, column, method = "iqr", threshold = 1.5)

Arguments

Value

A data frame with the modified column values.

Examples

# Example: 100 is an outlier
df <- data.frame(val = c(1, 2, 3, 2, 1, 100))

# The 100 will be replaced by the maximum allowed IQR value
clean_df <- treat_outliers(df, "val", method = "iqr")
print(clean_df$val)