Fine-tuned DistilBERT model for classifying perceived social support for healthy eating on a 1-5 scale. Part of a Theory of Planned Behavior (TPB) inference system for nutrition coaching.

This model is a fine-tuned version of distilbert-base-uncased for behavior change inference in nutrition coaching contexts.

Training Data

Training samples: 175
Validation samples: 25
Test samples: 50
Total: 250 samples

Performance

Test Accuracy: 78.0%

Intended Use

This model is designed for:

Nutrition coaching chatbots
Behavior change interventions
Health psychology research
Personalized dietary guidance

How to Use

python

from transformers import DistilBertTokenizer, DistilBertForSequenceClassification
import torch

# Load model
tokenizer = DistilBertTokenizer.from_pretrained("franzzzzzzzzz/tpb-subjective-norm-nutrition")
model = DistilBertForSequenceClassification.from_pretrained("franzzzzzzzzz/tpb-subjective-norm-nutrition")

# Predict
text = "I love healthy food, it's amazing!"
inputs = tokenizer(text, return_tensors='pt', padding=True, truncation=True)
outputs = model(**inputs)
prediction = torch.argmax(outputs.logits, dim=1).item()

# Convert 0-4 to 1-5 scale
score = prediction + 1
print(f"Score: {score}/5")

Limitations

Trained on English text only
Limited to nutrition/dietary contexts
May not generalize to other health behaviors
Requires context-appropriate input

Citation

If you use this model, please cite:

@misc{tpb-ttm-nutrition-models,
  author = {Your Name},
  title = {franzzzzzzzzz/tpb-subjective-norm-nutrition},
  year = {2026},
  publisher = {Hugging Face},
  howpublished = {\url{https://huggingface.co/franzzzzzzzzz/tpb-subjective-norm-nutrition}}
}

License

MIT License