Mollita/SecondaryModel

该数据集名为“SecondaryModel — 加密货币预测的元标记数据集”，旨在支持金融预测的可靠性改进研究。它包含用于完整复现实验的所有必要数据：针对20种加密货币资产和10个时间粒度，预处理的M1预测CSV文件，覆盖四个金融基础模型（Chronos-2、Fincast、Kronos、TiRex）；还包括跨资产外部特征序列（如恐惧贪婪指数、隐含波动率和新闻情感数据），以及预构建的PyTorch数据集缓存，这些缓存可以跳过多小时的数据组装步骤，直接用于实验。数据集适用于表格分类和时间序列预测任务，专注于金融、加密货币、元标记、选择性分类和算法交易领域。This dataset, named SecondaryModel — Meta-Labeling Dataset for Crypto Forecasting, is designed to support research on improving the reliability of financial forecasting. It contains all the data required to fully reproduce experiments: pre-processed M1 prediction CSV files for four financial foundation models across 20 crypto assets and 10 granularities, cross-asset external feature series (e.g., Fear & Greed Index, implied volatility, and news sentiment data), and pre-built PyTorch dataset caches that skip the multi-hour data-assembly step. The dataset is suitable for tabular classification and time-series forecasting tasks, focusing on finance, crypto, meta-labeling, selective classification, and algorithmic trading domains.