Eco-Amazon: Enriching E-commerce Datasets with Product Carbon Footprint for Sustainable Recommendations

This repository contains the Amazon datasets enriched with Product Carbon Footprint (PCF).

Such dataset have been obtained by prompting state-of-the-art Large Language Models (LLMs) for estimating the PCF of Amazon products, in the Clothing, Electronics, Home & Kitchen domains. In particular, we exploit Google Gemini 2.5 Flash and OpenAI o3-mini to infer CO2e emissions based on product metadata, following strictly defined Life Cycle Assessment (LCA) standards (GHG Protocol, ISO 14040/14044).

More details about the way these datasets have been enriched can be found in the associated paper and our repository for the source code.

Metric
Electronics
Home & Kitchen
Clothing

Total Users
21,751
66,810
97,608

Total Items
11,495
17,027
21,380

Total Ratings
464,464
684,651
1,070,586

We provide these datasets in two forms:

item metadata enriched with PCF estimations, one per LLM, in json format. For example, these are the Clothing datasets enriched with PCF estimations provided by Gemini-2.5-flash and GPT-o3-mini:

clothing_gemini.jsonlclothing_o3mini.jsonl

datasets in the RecBole format, used in our use case, whose code can be found in our GitHub Repository. As an example, the Amazon Clothing dataset in the RecBole format is the following:

amazon_clothing.zip