FoodNetBD: A Bilingual, Disease-Annotated Nutritional Dataset for Bangladesh

Description

FoodNetBD is an openly licensed, machine-readable nutritional dataset of 600 food items commonly consumed across the Bangladeshi diet. Each item is described by 33 nutrient variables per 100 g edible portion, a bilingual English–Bengali (Unicode) name, and a structured Disease_Tags field assigning the item to one of three constraint tiers (safe, restrict, avoid) across twelve diet-sensitive clinical conditions (type 2 diabetes, PCOS, hypertension, hypotension, chronic kidney disease, anemia, gout, irritable bowel syndrome, asthma, hypothyroidism, hyperthyroidism, and celiac disease). Nutrient values were compiled from the Food Composition Table for Bangladesh, the Indian Food Composition Tables 2017, USDA FoodData Central, and supplementary national and regional composition databases, with every row carrying source citation, source-type, data-completeness, and energy-calculation-method fields. The dataset is directly reusable for disease-aware meal-planning and clinical decision-support tools, Bengali-language health applications, algorithmic benchmarking on tabular nutritional data, and epidemiological research on the Bangladeshi diet. File: - FoodNetBD.csv — the primary dataset (600 rows × 56 columns)

Steps to reproduce

Select 600 commonly consumed foods from national dietary surveys, covering all 16 food groups and nutritionally distinct preparation variants (e.g., raw, boiled, dried, cooked). Compile nutrient values using a source hierarchy: FCTB, IFCT 2017, USDA FoodData Central, and FAO/INFOODS. Use values closest to Bangladeshi preparation methods. Indigenous species are curated from analytical studies and labeled Analytical; others are Compiled. Record energy in kcal and kJ, classify the calculation method (Atwater, Modified Atwater, or Direct), and compute Vitamin A as retinol activity equivalents following FAO/INFOODS guidelines. Assign English and Bengali (Unicode) names to every food, distinguishing preparation states and regional varieties where nutritionally relevant. Assign Disease_Tags (safe/restrict/avoid) for 12 conditions using published nutrient thresholds and clinical guidelines. Record provenance for each item, including source citation, source type, completeness score, energy calculation method, and proximate-sum note. Validate the dataset through internal consistency, energy and proximate audits, external cross-source comparison, statistical testing, and machine learning label validation. Release the validated dataset as a UTF-8 CSV (600 × 56) under a CC BY 4.0 license.

Institutions

• Khulna University of Engineering and Technology

  Khulna Division, Khulna

Categories

Licence