| # CryptoTrendPredictor | |
| ## Overview | |
| CryptoTrendPredictor is a BERT-based model fine-tuned to predict short-term Bitcoin price movement direction (up or down) within the next 24 hours based on crypto-related news headlines and social media text. It outputs a binary prediction: **up** or **down**. | |
| This model is intended for research and educational purposes only. It is not financial advice. | |
| ## Model Architecture | |
| - **Base model**: BERT-base-uncased | |
| - **Task head**: Binary classification head | |
| - **Hidden size**: 768 | |
| - **Number of layers**: 12 | |
| - **Parameters**: ~110M | |
| Built using `BertForSequenceClassification` from the Transformers library. | |
| ## Usage | |
| ```python | |
| from transformers import pipeline | |
| predictor = pipeline( | |
| "text-classification", | |
| model="your-username/CryptoTrendPredictor" | |
| ) | |
| text = "Bitcoin ETF approved by SEC, major institutions entering market" | |
| result = predictor(text) | |
| print(result) | |
| # [{'label': 'up', 'score': 0.92}] |