Chinese Financial Sentiment Analysis Model (Crypto Focus)

中文金融情感分析模型(加密货币领域)

模型描述 | Model Description

本模型基于 yiyanghkust/finbert-tone-chinese 微调,专门用于分析中文加密货币相关新闻和社交媒体内容的情感倾向。模型可以识别三种情感类别:正面(Positive)、中性(Neutral)和负面(Negative)。

This model is fine-tuned from yiyanghkust/finbert-tone-chinese and specifically designed for sentiment analysis of Chinese cryptocurrency-related news and social media content. It can classify text into three sentiment categories: Positive, Neutral, and Negative.

训练数据 | Training Data

  • 数据量 | Size: 1000条人工标注的中文金融新闻 | 1000 manually annotated Chinese financial news articles
  • 数据来源 | Source: 加密货币相关新闻和推文 | Cryptocurrency-related news and tweets
  • 标注方式 | Annotation: AI辅助 + 人工修正 | AI-assisted + Manual correction
  • 数据分布 | Distribution:
    • Positive(正面): 420条 (42.0%)
    • Neutral(中性): 420条 (42.0%)
    • Negative(负面): 160条 (16.0%)

性能指标 | Performance Metrics

在200条测试集上的表现 | Performance on 200 test samples:

指标 Metric 数值 Value
准确率 Accuracy 64.50%
F1分数 F1 Score 63.65%
精确率 Precision 63.94%
召回率 Recall 64.50%

使用方法 | Usage

快速开始 | Quick Start

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

# 加载模型和分词器 | Load model and tokenizer
model_name = "LocalOptimum/chinese-crypto-sentiment"  
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForSequenceClassification.from_pretrained(model_name)

# 分析文本 | Analyze text
text = "比特币突破10万美元创历史新高"
inputs = tokenizer(text, return_tensors="pt", truncation=True, max_length=128)

# 预测 | Predict
with torch.no_grad():
    outputs = model(**inputs)
    predictions = torch.nn.functional.softmax(outputs.logits, dim=-1)
    predicted_class = torch.argmax(predictions, dim=-1).item()

# 结果映射 | Result mapping
labels = ['positive', 'neutral', 'negative']
sentiment = labels[predicted_class]
confidence = predictions[0][predicted_class].item()

print(f"情感: {sentiment}")
print(f"置信度: {confidence:.4f}")

批量处理 | Batch Processing

texts = [
    "币安获得阿布扎比监管授权",
    "以太坊完成Fusaka升级",
    "某交易所遭攻击损失100万美元"
]

inputs = tokenizer(texts, return_tensors="pt", truncation=True,
                   max_length=128, padding=True)

with torch.no_grad():
    outputs = model(**inputs)
    predictions = torch.nn.functional.softmax(outputs.logits, dim=-1)
    predicted_classes = torch.argmax(predictions, dim=-1)

labels = ['positive', 'neutral', 'negative']
for text, pred in zip(texts, predicted_classes):
    print(f"{text} -> {labels[pred]}")

训练参数 | Training Configuration

  • 基础模型 | Base Model: yiyanghkust/finbert-tone-chinese
  • 训练轮数 | Epochs: 5
  • 批次大小 | Batch Size: 16
  • 学习率 | Learning Rate: 2e-5
  • 最大序列长度 | Max Length: 128
  • 训练设备 | Device: NVIDIA GeForce RTX 3060 Laptop GPU
  • 训练时间 | Training Time: ~5分钟 | ~5 minutes

适用场景 | Use Cases

  • ✅ 加密货币新闻情感分析
  • ✅ 社交媒体舆情监控
  • ✅ 金融市场情绪指标
  • ✅ 实时新闻情感跟踪
  • ✅ 投资决策辅助参考

局限性 | Limitations

  • ⚠️ 主要针对加密货币领域的金融新闻,其他金融领域可能表现不佳
  • ⚠️ 负面样本相对较少(16%),对负面情感的识别可能不够敏感
  • ⚠️ 短文本(少于10字)的分析准确率可能下降
  • ⚠️ 仅支持简体中文
  • ⚠️ 模型不能替代人工判断,仅供参考

许可证 | License

Apache-2.0

引用 | Citation

如果使用本模型,请引用:

@misc{watchtower-sentiment-2025,
  title={Chinese Financial Sentiment Analysis Model (Crypto Focus)},
  author={Onefly},
  year={2025},
  howpublished={\url{https://huggingface.co/YOUR_USERNAME/sentiment-finetuned-1000}},
  note={Fine-tuned from yiyanghkust/finbert-tone-chinese}
}

基础模型 | Base Model

本模型基于以下模型微调:

感谢原作者的贡献!

更新日志 | Changelog

v2.0 (2025-12-09)

  • ✅ 扩充训练数据至1000条
  • ✅ 修正标注错误,提升数据质量
  • ✅ 优化类别分布,提升模型平衡性
  • ✅ F1分数提升2.01%(0.6165 → 0.6365)

v1.0 (Initial Release)

  • 基于500条标注数据的初始版本

联系方式 | Contact

如有问题或建议,欢迎提 issue 或 PR。


维护者 | Maintainer: Onefly 最后更新 | Last Updated: 2025-12-09

Downloads last month
14
Safetensors
Model size
0.1B params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Evaluation results