Skip to main content

Mangetamain - Recipe Clustering & Exploration

A data-driven recipe exploration platform combining clustering, nutritional analysis, and interactive visualization through a reproducible Streamlit pipeline.

Visit repository
  • Data Analysis
  • Machine Learning
  • Feature Engineering
  • Streamlit App Development
  • CI/CD & Reproducible Pipelines
Mangetamain recipe exploration platform.

A platform for recipe exploration

Mangetamain was designed as an interactive platform for exploring large-scale recipe data through a user-friendly web interface. The goal was to go beyond simple recipe browsing and provide a structured way to analyze nutritional composition, sensory patterns, preparation habits, and user interactions.

Mangetamain interface.
Feature engineering methodology.

From raw data to meaningful features

A central part of the project was building a feature engineering pipeline capable of transforming raw recipe and rating data into meaningful analytical signals. This included nutritional indicators, ingredient and step counts, interaction-based metrics, temporal patterns, and confidence-aware aggregations.

Clustering recipes into interpretable groups

One of the core objectives of Mangetamain was to identify coherent groups of recipes based on nutritional and behavioral characteristics. Instead of relying only on cuisine labels or manually defined categories, the project used clustering techniques to uncover patterns directly from the data.

This approach made it possible to highlight distinct recipe profiles, such as lighter or more energy-dense dishes, recipes with shorter preparation times, or clusters shaped by stronger user engagement. The result is a more data-driven way of navigating culinary diversity.

Recipe clustering visualization.

Interactive analysis at scale

The final application combines a reproducible preprocessing pipeline, clustering workflows, and an interactive Streamlit interface to make large recipe datasets accessible and interpretable.

Users can explore distributions, compare clusters, inspect ingredients and nutritional patterns, and better understand how recipes are organized across multiple dimensions. Beyond the web app itself, the project also emphasized maintainability and reproducibility through modular architecture, testing, Docker support, automated documentation, and a clean development workflow.