kbatyshchev/thesis-llm-financial-forecaster

Multimodal Price Signal Predictor

This model predicts a trading signal (“Buy”, “Hold”, “Sell”) for selected stocks and cryptocurrencies by directly leveraging both technical indicators and raw news headlines as inputs.

Input:

• Technical/time-series features
• Raw news headline text

Output:

• A single predicted return value 1

Signal Interpretation:

• Buy: if return > 1.03
• Sell: if return < 0.97
• Hold: otherwise

Datasets Used:

• News data: https://www.kaggle.com/datasets/parzik/news-thesis
• Time series: https://www.kaggle.com/datasets/parzik/thesis-timeseries-ready-to-use

Example output:

Output return: 0.94
→ Action: sell

Notes:

• If something does not work, the changes can be demanded from k.batyshchev@innopolis.university