README.md

# 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}]