dataframe - 如何处理不同频率的数据集？(How to process dataset to different frequency?)

Question

I built a model to predict user activity using accelerometer data and I used a dataset which samples where collected at 100 Hz.

(我建立了一个使用加速度计数据预测用户活动的模型，并使用了一个以100 Hz采样的数据集。)

What can I do with this data to customize the model for predicting samples that will be collected at 5 Hz?

(如何处理这些数据以定制模型以预测将以5 Hz频率采集的样本？)

Are there any good ways to regularize in this case?

(在这种情况下，有什么好的方法可以规范化吗？)

I built the 1D CNN model using the keras library, I've load my data in dataframe,

(我使用keras库构建了1D CNN模型，并将数据加载到数据框中，)

  ask by mayerpaw translate from so

Answer 1

I think the only way to achieve this is to modify your training data if this is a possibility for you.

(我认为实现此目标的唯一方法是修改训练数据（如果您有可能的话）。)

Instead of taking the 100 Hz data directly, you just use every 20th frame of it.

(您无需直接获取100 Hz数据，而只需使用每20帧。)

But you could perform some data augmentation by splitting the data into multiple sets by introducing a sample offset of 0 to 19. Otherwise, you have to modify your 5 Hz data by up-sampling it to 100 Hz.

(但是您可以通过引入0到19的样本偏移量将数据分成多个集合来执行一些数据增强。否则，您必须通过将其上采样到100 Hz来修改5 Hz数据。)

This would need some kind of interpolation between the data points which might be not that easy and introduces an error.

(这将需要在数据点之间进行某种类型的内插，这可能不是那么容易，并且会引入错误。)