Green AI Product Recommender System

CHAPTER 1
INTRODUCTION

1.1 Background

With rapid growth in industries and consumerism, people now have access to
thousands of products in every category such as food, beverages, hygiene, and
household essentials. While this gives freedom of choice, it also leads to an increase
in products that harm the environment due to high carbon footprint, plastic packaging,
chemical use, and unsustainable manufacturing methods. As environmental awareness
rises, users want to make eco-friendly choices — but they lack proper information to
compare products based on sustainability. Green AI Product Recommender System
is developed to bridge this gap by recommending greener alternatives using artificial
intelligence, helping individuals choose products that are better for the environment.
The system reads product information, analyzes environmental scores, and suggests
more sustainable options automatically. By combining AI with sustainability goals,
the project supports Greener Technology Adoption, consumer awareness, and the
United Nations Sustainable Development Goals (SDGs) related to responsible
consumption.

1.2 Problem Statement

Choosing eco-friendly products is difficult because:
 Most users are unaware of the environmental impact of everyday products.
 Many products lack clear sustainability scores on packaging.
 Searching manually for greener alternatives is time-consuming.
 Existing platforms like OpenFoodFacts provide data but not intelligent
suggestions.
Therefore, there is a need for an AI-based system that helps users quickly search
products and discover better, environment-friendly options in real-time.

Dept of CSE, EPCET 2025-2026

 Green AI Product Recommender System

1.3 Existing System

Existing systems majorly provide only product information.

Examples:
 Online stores → show product details but no eco suggestion

 Sustainability APIs → give eco-scores but no personalization

 Manual comparison → slow and unreliable

Limitations of existing solutions:

Limitation Impact
No recommendation Users cannot find greener
engine alternatives easily
Requires manual product
No automation
search
No focus on sustainability
Not user-centric
decision-making
Raw data not useful for
Limited accessibility
general consumers

Hence, existing systems fail to guide users toward better purchasing decisions.

1.4 Proposed System (Mention Which Software Model is Used)

The proposed Green AI Product Recommender intelligently searches and recommends
products based on:
 Exact search
 Fuzzy matching for name variations
 Semantic AI model (Transformer-based embeddings)
 Rule-based eco-friendly suggestions

Key features:
 Unified Smart Search — finds best match from database using AI
 Eco-score comparison — supports informed decisions
 Sustainable alternative suggestions
 User-friendly web interface
 Automatic product imports from open data sources

This motivates environment-friendly shopping behaviour.

Dept of CSE, EPCET 2025-2026

 Green AI Product Recommender System

1.5 Aim of the Project

The aim of this project is to develop an intelligent Green AI Product Recommender

System that suggests eco-friendly product alternatives using AI-based search and
sustainability scoring. The system enhances user awareness by identifying greener
options based on factors such as organic ingredients, reduced environmental footprint,
and sustainable sourcing. By integrating fuzzy matching, semantic similarity, and
rule-based AI, the system ensures accurate recommendations even when product names
are partially matched or misspelled. It uses open sustainability datasets to provide
reliable and updated results. Ultimately, this project encourages consumers to make
environmentally responsible choices while offering a fast, user-friendly experience for
discovering greener product alternatives.

1.6 Objective of the project

 Enable Real-Time Green Product Search: Allow users to quickly search for
consumer products using text-based queries.
 Suggest Eco-Friendly Alternatives: Recommend greener products based on
sustainability factors such as organic sourcing and reduced environmental
impact.
 Utilize AI-Driven Techniques: Implement semantic similarity, fuzzy
matching, and rule-based AI to improve search accuracy and alternative
identification.
 Support Continuous Dataset Updates: Automatically import fresh
information from open sustainability data sources to ensure reliable results.
 Increase User Awareness: Provide clear explanations on why an eco-friendly
alternative is better, promoting responsible consumer choices.
 Focus on Sustainability Metrics: Use eco-scores and category-based analysis
to compare products and highlight environmentally safer options.

Dept of CSE, EPCET 2025-2026

 Green AI Product Recommender System

CHAPTER 2

LITERATURE SURVEY

“Eco-Friendly Product Recommendation System Using Machine Learning Techniques” – 2022

Abstract: With the rising need for environmental responsibility, recommender systems are
being adapted to suggest sustainable alternatives. The system analyzes product attributes like
carbon footprint, recyclability, and eco-standards to compute a sustainability score. It applies
machine learning classification and similarity matching to recommend greener substitutes.

Pros: Helps consumers make environmentally responsible decisions. Automated knowledge

extraction from large datasets.

Cons: Eco-score availability is limited for many products, leading to partial recommendations.

“Green Consumer Awareness Using Sustainable Product Databases” – 2023

Abstract: The study emphasizes integrating open eco-labeled datasets to improve

recommendation accuracy and transparency. Databases like OpenFoodFacts enable users to
view product composition, grade levels, and certifications for informed decision-making.

Pros: Open-source and widely accessible product data. Increased transparency and trust.

Cons: Data inconsistencies in publicly contributed datasets require heavy preprocessing.

“AI-Based Product Retrieval Using Semantic and Fuzzy Search” – 2021

Abstract: Focused on enhancing search results using NLP-based fuzzy matching and semantic
vector similarity. User queries are compared with product titles using cosine similarity and
contextual embeddings to retrieve more relevant results.

Pros: Improves search quality even with spelling errors and incomplete queries.

Cons: Semantic models can be computationally costly for large-scale datasets.

Dept of CSE, EPCET 2025-2026

 Green AI Product Recommender System

“Sustainable Recommendation System Using Nutri-Score and Eco-Score Integration” – 2020

Abstract: Introduces dual scoring by combining nutritional value and environmental impact to
rank food products. It suggests better choices promoting both user health and ecological
benefits.

Pros: Multi-criteria scoring leads to more meaningful recommendations.

Cons: Eco-score data not yet fully available for many commercial products.

“Automated Green Product Suggestion using Knowledge-Based AI Rules” – 2024

Abstract: Proposes rule-based logic to intelligently suggest greener alternative products where
data-driven matches fail. It uses predefined sustainability patterns such as “replace plastic with
biodegradable” or “prefer organic over conventional”.

Pros: Works even when dataset similarity is low, improving coverage.

Cons: Rules need continuous updates to reflect new product trends and sustainability policies.

“Carbon Footprint-Aware Recommender System for Consumer Goods” – 2023

Abstract: This research focuses on integrating carbon emission indicators into product
recommendation engines. Each product is assigned an estimated carbon footprint based on
sourcing, packaging, and transportation data. Recommendations prioritize products with lower
environmental impact while still maintaining quality preferences.

Pros: Promotes climate-conscious decision-making and quantifiable sustainability improvements.

Cons: Carbon footprint data varies across regions and supply chains, making standardization
difficult.

Dept of CSE, EPCET 2025-2026

 Green AI Product Recommender System

CHAPTER 3

REQUIREMENTS SPECIFICATIONS

3.1 Hardware Requirements

i. Processor: Modern multi-core processor (Intel i5/i7 or AMD equivalent).

ii. Memory (RAM): Minimum 8 GB (16 GB recommended for AI model execution).

iii. Storage: At least 50 GB free space for project files, embeddings & local DB (if used).

iv. Internet Connection: Stable broadband internet for API communication, dataset imports
& live updates.

Software Requirements
i. Operating System: Windows 10/11, macOS, or Linux.

ii. Programming Languages:

• Python 3.10 or above (Backend & AI)
• JavaScript (Frontend)

iii. Frameworks & Libraries:

• Flask — Backend API Framework
• RapidFuzz — Fuzzy Search Matching
• Sentence Transformers — Semantic Search
• PyMongo — MongoDB connectivity

iv. Database: MongoDB Atlas (Cloud) or Local MongoDB Server

v. Frontend Requirements:
• HTML, CSS, Tailwind CSS, JavaScript
• Modern browsers (Chrome, Firefox, Edge, Safari)

vi. Development Tools:

• VS Code / PyCharm
• Postman for API testing
• GitHub for version control

vii. APIs & External Data Sources:

• OpenFoodFacts — Product Data
• Internet connectivity for Live Import Updates
• Kaggle datasets (optional)

Dept of CSE, EPCET 2025-2026

 Green AI Product Recommender System

CHAPTER 4

SYSTEM DESIGN

4.1 System Architecture

The system architecture for the Green AI Product Recommender System includes multiple
integrated components that ensure accurate product searches and sustainability-based
[Link] interact with the web interface by entering a product name.
The backend receives the query and processes it using multiple AI search techniques such as:
 Exact Text Search
 Fuzzy Matching
 Semantic Similarity (MiniLM Transformer Model)
 Rule-based Eco Alternatives
The MongoDB database stores product information including eco-scores and sustainability
descriptions.
Live data import functions continuously update the database with the latest eco-friendly
product data from open sources such as OpenFoodFacts.
This architecture ensures high search reliability, real-time updates, and environmental
awareness, ultimately helping users choose greener products easily and efficiently.

Dept of CSE, EPCET 2025-2026

 Green AI Product Recommender System

4.2 Software Model

The proposed system follows the MVC (Model-View-Controller) architecture:

Layer Component Purpose
MongoDB Database Stores product data and performs
Model
+ AI Models semantic & fuzzy similarity search
Frontend (HTML, Displays product alternatives and
View
CSS, JavaScript) search results to the user
Controls logic and communication
Controller Flask Backend API
between UI and Database

Workflow Explanation:
1) User enters a product name in UI
2) Controller sends request to backend API
3) Model fetches & processes results using:
 Regex filtering
 RapidFuzz scoring
 Transformer-based sentence embeddings
4) Best matches are ranked and returned
5) View displays eco-friendly product suggestions with eco score and reason for
recommendation
6) System improves over time with updated green datasets

Dept of CSE, EPCET 2025-2026

 Green AI Product Recommender System

4.3 Data Flow Model

The data flow for the system includes the following steps:
1. User Interaction: User searches for a product through the search interface
2. Input Processing: Query sent to server for interpretation and similarity assessment
3. Recommendation Generation: AI selects the best eco-friendly match using combined
algorithms
4. Database Access: MongoDB returns stored sustainability details
5. Display Results: Recommendations shown with eco score + why it is better

Dept of CSE, EPCET 2025-2026

 Green AI Product Recommender System

CHAPTER 5

SYSTEM IMPLEMENTATION

5.1 Procedure to prepare model

The procedure to prepare the Green AI Product Recommender System model involves several
systematic stages, starting from dataset acquisition to AI-driven recommendation generation. The
objective is to ensure accurate, efficient, and eco-focused product suggestions. The steps involved are as
follows:

➤Data Collection and Integration: Gather product data from various open-source
sustainability datasets such as OpenFoodFacts, OpenEcoScore, EcoLabel Index, and
GreenDB. Consolidate attributes like product name, category, eco-score, ingredients,
sustainability labels, and environmental certifications. Clean and preprocess the data by
handling missing values, resolving duplicates, and standardizing formats.

➤ AI Search Pipeline Development: Design the intelligent search workflow by

integrating multiple techniques—Exact Match, Fuzzy Matching, Semantic Embedding
Similarity, and Rule-Based Green AI logic. This multi-layered pipeline ensures that even
imperfect or unknown user inputs yield accurate eco-friendly recommendations.

➤ Model Preparation and Embedding Generation: Use sentence-transformer models to

convert product names and user queries into semantic embeddings. Generate vector
representations for all products in the database. This enables deeper understanding of user
intent and supports semantic similarity computation for green alternatives.

➤ Rule-Based Sustainability Engine: Define environmentally focused rules for

suggesting greener alternatives. These rules consider eco-scores, organic labeling,
environmental certifications, and product categories. The engine ensures that the system
consistently recommends products with lower environmental impact.

Dept of CSE, EPCET 2025-2026

 Green AI Product Recommender System

➤ Backend Integration: Implement the smart search pipeline within the Flask backend.
Connect the logic to the MongoDB database for real-time retrieval of product details, eco
scores, and alternative options. Validate the end-to-end data flow from user query to AI-
based output.

➤ Testing and Refinement: Evaluate search accuracy, semantic relevance, and

sustainability correctness. Fine-tune similarity thresholds, fuzzy scores, and rule-based
logic. Ensure the system handles typos, incomplete names, and unknown queries
gracefully.

➤ Deployment: Integrate the backend and frontend to enable seamless real-time product
search. Schedule automatic updates to refresh datasets weekly, ensuring the model remains
accurate with the latest eco-friendly product information.

Dept of CSE, EPCET 2025-2026

 Green AI Product Recommender System

5.2 Code Snippet

Dept of CSE, EPCET 2025-2026

 Green AI Product Recommender System

CHAPTER 6

RESULT & ANALYSIS

6.1 Snapshots

Fig : Home page of website

Fig : Search Results for a product

Dept of CSE, EPCET 2025-2026

 Green AI Product Recommender System

6.2 Result analysis

Result analysis evaluates the performance, effectiveness, and reliability of the Green AI Product
Recommender System. The system was assessed based on multiple key parameters that determine the
quality of recommendations, efficiency of search mechanisms, and user satisfaction. The major findings
are summarized below:

1. Recommendation Accuracy: The system demonstrates high accuracy by integrating

exact match, fuzzy matching, and semantic similarity using transformer embeddings.
Cosine similarity scores were used to evaluate how closely the identified eco-friendly
alternatives align with the user’s query. This resulted in significantly precise and
meaningful recommendations.

2. Search Efficiency: The unified smart search pipeline ensures fast response times, even
for complex or ambiguous queries. Fuzzy logic effectively handles typos, partial
matches, and spelling variations, while semantic search improves contextual
understanding of user inputs.

3. Sustainability Awareness & Insights: The system successfully highlights the eco-score
and why the suggested alternative is better, helping users make informed,
environmentally responsible decisions. This feature increases user awareness of
product sustainability and encourages greener choices.

4. Recommendation Diversity: The model ensures that the suggested products are not
repetitive by considering alternative scores, categories, and sustainability indicators.
This improves the diversity and novelty of generated recommendations.

5. Cold Start Handling: In cases where products are not found in the database, the rule-
based AI module generates meaningful fallback eco-friendly suggestions. This ensures
system functioning even with limited or missing data.

6. User Interaction & Engagement: Feedback from testing showed that users found the
interface intuitive, with clear recommendations and explanations. The interactive UI
combined with instant responses increased user satisfaction and engagement levels.

Dept of CSE, EPCET 2025-2026

 Green AI Product Recommender System

CHAPTER 7

CONCLUSION

The Green AI Product Recommender System was successfully designed and developed to promote
sustainable consumer choices using AI-driven search and intelligent product analysis. The system
integrates exact matching, fuzzy logic, semantic similarity, and rule-based AI to deliver accurate
and eco-friendly product suggestions. By combining multiple open-source datasets and
sustainability indicators such as eco-scores, the system provides users with meaningful insights into
why a recommended alternative is greener and more responsible.

The project effectively addresses the challenge of identifying sustainable product options in a
rapidly growing market with diverse and often confusing environmental claims. Through its smart
search pipeline, intuitive interface, and dynamic recommendation engine, the system enhances user
awareness and simplifies the decision-making process. The inclusion of automated dataset updates
further ensures that the recommendations remain relevant and up to date.

Overall, this system demonstrates how AI can be leveraged to support environmental sustainability,
guiding users toward eco-conscious decisions and contributing to greener consumption patterns.
Future improvements may include expanding datasets, integrating real-time APIs, adding
personalized recommendations using user profiles, and incorporating deeper LLM-based reasoning
for sustainability evaluation.

Dept of CSE, EPCET 2025-2026

 Green AI Product Recommender System

REFERENCES

[1] OpenFoodFacts. (2024). Global Open Food Products Database — sustainability & eco-score
metadata.

[2] OpenEcoScore Team. (2023). Eco-Scoring Framework for Environmental Impact Analysis of
Consumer Products.

[3] Digital Green Organization. (2022). GreenDB: Open Database of Eco-Friendly Product
Information.

[4] EcoLabel Index. (2023). Global Registry of Environmental Certifications & Eco-Labels.

[5] Reimers, N., & Gurevych, I. (2019). Sentence-BERT: Semantic similarity using transformer
models.

[6] Bachmann, M. (2023). RapidFuzz: Fast Fuzzy String Matching Library for Python.

[7] MongoDB Inc. (2024). MongoDB NoSQL Database Documentation — scalable product storage.

[8] Flask Developers. (2024). Flask: Lightweight Web Framework for Python-Based Applications.

[9] Hugging Face. (2024). Transformer Models for Natural Language Processing and Embedding
Generation.

[10] Kaggle Datasets. (2024). Sustainable Consumer Products & Environmental Impact Datasets.

Dept of CSE, EPCET 2025-2026