This project implements a product recommendation algorithm and carries out customer analysis using LASSO regression.

The objectives of the project are:

• Develop a set of product recommendation rules, based on the Apriori algorithm
• Rank customers on the basis of their marginal contribution to revenues.

The ultimate goal is to provide information that can be used by a marketing department to develop more targeted campaigns.

The core of the analysis is contained in these three files:

• Customers_by_product.sql
• data_and_analysis.php
• analysis.R

Note that some of the key SQL queries, to generate the data for the analysis but also to create a network graph, are contained in both the R file and the additional Customers_by_product.sql file. The latter is called by the setup script after the database is populated.

To develop the product recommendation system we have used the Apriori algorithm. We provide a link to the relevant Wikipedia article on the introductory page of the web application.

To develop the LASSO regression, we first have identified the top 20 customers in terms of total revenues generated. Then we have carried out a LASSO regression using the lars package with the objective to narrow down the number of customers with a significant marginal contribution to revenues.

The 'Data' tab includes a network graph of the links between product categories. Note that the graph is generated dynamically each time the ./setup.sh run command is given. The script saves a .png file in the /web sub-directory, which is then retrieved via html link.

The R analysis relies on the following packages.

• igraph
• arules
• lars

This project is based on code by: Guglielmo Bartolozzi, Gaston Besanson, Christian Brownlees, Stefano Costantini, Laura Cozma, Jordi Zamora Munt