← 返回刚才阅读的位置

知识卡片

普通读书笔记卡 · 6488.u · 生成模型:GPT-6

时间分箱权衡分辨率与稀疏

理解等级:中级

把每次独立时间戳归入时段、星期或季节,可减少情境组合数量并增加每格记录。分箱越粗越可能合并不同行为,越细越可能样本不足,应按任务选择并保持训练与预测编码一致。

这张卡片看不懂?

已有反馈 · 0 条

    还没有人提交反馈。

    在整书笔记中的位置

    原书出处 · 1 处

    • · 查看原书依据

      the check-in timestamp is encoded into time slots representing a season, day of the week, and hour slot, ensuring dimension reduction. Usually, different users mostly prefer visiting contrasting locations in distinct seasons of the year. As a result, locations also experience varying check-in distribution. Not only seasons but users also prefer visiting places at different hours of the day. Hence, the entire day is divided

    参考来源

    - 位置:《Artificial Intelligence and Data Science in Recommendation System: Current Trends, Technologies, and Applications》第11章《张量分解的兴趣点推荐》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Artificial Intelligence and Data Science in Recommendation System: Current Trends, Technologies, and Applications/_epub-src/OEBPS/Text/9789815136746-C11.html) - 结论依据:摘录中关于时间分箱权衡分辨率与稀疏的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。 - 原始内容:the check-in timestamp is encoded into time slots representing a season, day of the week, and hour slot, ensuring dimension reduction. Usually, different users mostly prefer visiting contrasting locations in distinct seasons of the year. As a result, locations also experience varying check-in distribution. Not only seasons but users also prefer visiting places at different hours of the day. Hence, the entire day is divided