how to reduce 10 data points to 8 proportionally - 未名空间MITBBS历史存档

国际科技财经博客移民网络热点娱乐民生时事公众号

Redian新闻

>未名空间

>DataSciences - 数据科学

how to reduce 10 data points to 8 proportionally

how to reduce 10 data points to 8 proportionally# DataSciences - 数据科学

s*t2017-05-08 07:05

1 楼

女生的套路到底有多深，我来带大家见识一下。我女朋友平时就经常查我的岗，有
一次她出差了，我晚上和朋友出去玩，12点还没回家。但机智如我在出门前就已经在家
里拍了各种照片、视频备用。
我：今天太累了先睡了哦宝贝，晚安。
她：你又去外面鬼混了吧？
我：没有啊，我在家呢
（发了她一张卧室的照片）
她：拍一段卫生间的视频给我看看
（还好我机智哈哈哈哈哈，转发了提前录好的小视屏）
她：好的，睡吧。哦对了，冰箱上面我放了100块钱，你把编号发给我一下吧。

y*32017-05-08 07:05

2 楼

Hello,
I am building a time series model. The first couple of years I have 8 data
points.
but this year I have 10 data points. I am trying to convert the 10 data
points to 8 proportionally so that I can use all they years data to catch
seasonality( the differences between two nearest neighbors within one year
are different) .
Is there a sound mathematical way to achieve this? Thanks

m*e2017-05-08 07:05

3 楼

男人的套路不深？

E*g2017-05-08 07:05

4 楼

interpolate to 40 points and pick one every 5.

y*32017-05-08 07:05

5 楼

thanks!!!!
planning to do linear interpolation. But what if the relationship is not
straightly linear? Linear interpolation is gonna affect the projection

E*g2017-05-08 07:05

6 楼

there are many ways to do interpolation. don't have
to be linear
https://en.wikipedia.org/wiki/Interpolation

y*32017-05-08 07:05

7 楼

Thanks!
My problem is I want the original 1st and 10th data points same and turn
the 8 data points in between into 6 data points. So when I apply
interpolation, am I going to only use the 1st and 10th to compute the linear
(if I choose linear interpolation) coefficient? and then compute the other
6?( the original 10 data points are equally spaced).

E*g2017-05-08 07:05

8 楼

1 to 10 : 9 spaces
=> 1 to 64: 63 spaces, pick 1 for every 9 points => get 8 points total,
including 1st and 10th
You'better find a library to do it in a non-linear way.
If using python, try:
https://docs.scipy.org/doc/scipy/reference/tutorial/interpolate.html

y*32017-05-08 07:05

9 楼

Great!!!thanks