收集并处理了13种手语数据集，用于双向手语翻译Collected and processed sign language datasets in 13 languages for bidirectional sign language translation
使用YouTube API和基于Python的自动化工具，从YouTube上收集了约2000个手语视频。这个GitHub仓库 展示了一个下载ASL（美国手语）视频的实例代码。Collected approximately 2000 sign language videos from YouTube using YouTube API and Python-based automation tools. This GitHub repository demonstrates example code for downloading ASL (American Sign Language) videos.
Jane Street 实时市场数据预测Jane Street Market Data Prediction
Kaggle（全球）Kaggle (Global)
团队成员，分析师Team Member, Analyst
2024年10月 - 2024年1月7日October 2024 - January 7, 2024
在全球3713支队伍中排名前4%（第143名），获得银牌Ranked top 4% (143rd) among 3,713 global teams, earned Silver Medal
主要成果Key Achievements
分析了包含47,127,338行样本的大规模数据集，涵盖1,699个交易日（超过四年半）的市场数据。数据集包含79个特征，代表不同金融指标。这个GitHub仓库 包含了完整的代码实现，包括特征工程、模型训练和优化过程。Analyzed a large-scale dataset containing 47,127,338 samples across 1,699 trading days (over 4.5 years). The dataset includes 79 features representing various financial indicators. This GitHub repository contains the complete code implementation, including feature engineering, model training, and optimization process.
应用K折交叉验证并优化LGBM模型，将预测准确率提高了10%Applied K-fold cross-validation and optimized LGBM model, improving prediction accuracy by 10%
通过超参数调优和稳健验证，将预测得分从0.0096提升至0.0083Enhanced prediction score from 0.0096 to 0.0083 through hyperparameter tuning and robust validation