Label Unbalance in High Frequency Trading Report: Label Unbalance in High-frequency Trading 高频交易中的标签不平衡问题研究 -- QuantML Method In this project, we mainly focus on the high frequency trading prediction in the scenario of label imbalance, based on machine learning methods. We provide four networks including: MLP, LSTM, BERT, and Mamba as the backbone. We use two simple methods to reduce the influence of label imbalance, including: resampling and class weighting. Dataset Due to copyright restrictions, we do not provide the original data. However, we provide the data structure and class distribution in our experiment. As a result, you can replace the dataset.py file with your own data and change some necessary parameters in trainclassification.py to run the code on your dataset. How to Run To run the model, you can use the following command: To change the model scale or other parameters related to training, please refer to the getargs function. We provide the following methods for addressing label imbalance: - --classweight: You can specify the weight for each class in the loss function.

⚠ No verified equity curve — no track-record source connected.

No score history is stored yet — only the current score is shown.

⚠ No live verification account connected — ask for proof before buying.

Alerts on changes: coming soon

Prop-firm compatibility not provided.

Data-completeness & trust index (not a profitability rating)