Introduction

This project demonstrates the application of association rule mining on the Online Retail Dataset. The goal is to uncover actionable insights for marketing, inventory, and sales strategies using association rules. The analysis includes data preparation, rule mining, and visualizations.

Step 1: Load Necessary Libraries

# Load required libraries
library(readr)
library(arules)
library(arulesViz)

Step 2: Load and Prepare the Data

# Load the dataset
online_retail <- read.csv("Online Retail.csv")

# Data Cleaning: Remove missing values and invalid entries
online_retail <- online_retail[!is.na(online_retail$InvoiceNo) & online_retail$Quantity > 0, ]

# Create a transactional dataset
data_transactions <- as(split(online_retail$Description, online_retail$InvoiceNo), "transactions")
summary(data_transactions)

Step 3: Exploratory Data Analysis

# Inspect transactions
inspect(data_transactions[1:5])

# Summary of frequent items
itemFrequencyPlot(data_transactions, topN = 20, col = "lightblue", main = "Top 20 Frequent Items")

Step 4: Generate Association Rules

# Apply the Apriori algorithm
rules <- apriori(data_transactions, parameter = list(supp = 0.01, conf = 0.5))

# Summary of rules
summary(rules)

# Filter top rules by confidence and lift
filtered_rules <- subset(rules, confidence > 0.7 & lift > 1.5)
summary(filtered_rules)

Step 5: Visualize Rules

# Scatter Plot
plot(filtered_rules, measure = c("support", "confidence"), shading = "lift", main = "Scatter plot of Filtered Rules")

# Two-Key Plot
plot(filtered_rules, method = "two-key plot", main = "Two-Key Plot of Filtered Rules")

# Interactive Graph
plot(filtered_rules, method = "graph", engine = "htmlwidget")

Step 6: Insights and Recommendations

# Insights:
# 1. High-confidence rules with strong lift indicate valuable item combinations for cross-selling.
# 2. Frequent item pairs (high support) suggest opportunities for bundle promotions.
#
# Recommendations:
# - Stock frequently associated items together to enhance customer convenience.
# - Use rules with medium confidence and high lift for discount bundling.

Conclusion

This analysis demonstrates the power of association rule mining for uncovering meaningful relationships in transaction data. The insights derived can guide effective marketing, inventory management, and sales strategies.