From Correlations to Recommendations

This repository contains materials for the Macroexpand 2025 conference presentation: "From Correlations to Recommendations – A Publisher's Journey into Data-Driven Book Sales using the SciCloj stack".

Overview

This presentation demonstrates how a book publisher (Jan Melvil Publishing) used Clojure and the SciCloj ecosystem to analyze customer purchasing patterns and develop a data-driven recommendation system.

Key Topics

• Discovering hidden patterns in customer purchasing behavior
• Moving from basic correlation analysis to association rule mining
• Building a practical book recommendation system
• Visualizing relationships between books using network graphs
• Implementing business insights from data analysis

Project Structure

The project uses a series of Clojure notebooks to demonstrate the journey from raw sales data to actionable recommendations:

1. Data loading and preparation
2. Correlation analysis between book titles
3. Market basket analysis using the Apriori algorithm
4. Recommendation engine implementation

Technical Stack

• Development Environment: VS Code + Calva

• Project Management: deps.edn

• Notebook Environment: Clay

• Key Libraries:

  • tablecloth.api - data manipulation
  • scicloj.tableplot - visualization
  • scicloj.kindly - rendering
  • clojure.math.combinatorics - itemset generation

Business Impact

This analysis led to several business improvements:

• More effective cross-selling recommendations
• Better understanding of purchase patterns between Czech vs foreign authors
• Creation of new targeted bundles based on actual customer behavior
• Implementation of "Customers Also Bought" sections on product pages

Resources

• Article about the topic on Clojure Civitas
• SciCloj: scicloj.github.io
• Tablecloth: scicloj.github.io/tablecloth

Contact

• Presenter: Tomáš Baránek
• About me: https://barys.me
• Email: tom@barys.me
• GitHub: https://www.github.com/tombarys
• Zulip: https://clojurians.zulipchat.com

Projects

• Jan Melvil Publishing: https://wwww.melvil.cz
• Servantes: Servant.es
• Blog (EN): https://lifehacky.net
• Blog and Club (CZ): https://lifehacky.cz
• Nautilus Omnibus: https://nautilus-omnibus.web.app