1

我有时间序列数据，我想执行傅立叶变换以确定所需时间序列的模式。对于每个时间间隔都有小波动，我想使用傅里叶变换来减少小波动的影响并确定该时间间隔的主要趋势。

数据样本：

 Date_And_Time ; Temperature
"2007-09-21 10:35:39";"7.480"
"2007-09-21 10:37:39";"7.500"
"2007-09-21 10:39:39";"7.510"
"2007-09-21 10:41:39";"7.520"
"2007-09-21 10:43:39";"7.440"
"2007-09-21 10:45:39";"7.470"
"2007-09-21 10:47:39";"7.560"
"2007-09-21 10:49:39";"7.740"
"2007-09-21 10:51:39";"7.570"
"2007-09-21 10:53:39";"7.730"
"2007-09-21 10:55:39";"7.740"
"2007-09-21 10:57:39";"7.930"
"2007-09-21 10:59:39";"7.870"
"2007-09-21 11:01:39";"7.870"
"2007-09-21 11:03:39";"7.680"
"2007-09-21 11:05:39";"8.160"
"2007-09-21 11:07:39";"8.260"
"2007-09-21 11:09:39";"8.020"
"2007-09-21 11:11:39";"7.900"
"2007-09-21 11:13:39";"7.940"
"2007-09-21 11:15:39";"7.910"
"2007-09-21 11:17:39";"8.040"
"2007-09-21 11:19:39";"8.120"
"2007-09-21 11:21:39";"8.090"
"2007-09-21 11:23:39";"8.140"
"2007-09-21 11:25:39";"8.230"
"2007-09-21 11:27:39";"8.200"
"2007-09-21 11:29:39";"8.380"
"2007-09-21 11:31:39";"8.240"
"2007-09-21 11:33:39";"8.250"
"2007-09-21 11:35:39";"8.340"
"2007-09-21 11:37:39";"8.350"
"2007-09-21 11:39:39";"8.350"
"2007-09-21 11:41:39";"8.350"
"2007-09-21 11:43:39";"8.360"
"2007-09-21 11:45:39";"8.490"
"2007-09-21 11:47:39";"8.560"
"2007-09-21 11:49:39";"8.370"
"2007-09-21 11:51:39";"8.080"
"2007-09-21 11:53:39";"8.150"
"2007-09-21 11:55:39";"8.360"
"2007-09-21 11:57:39";"8.410"
"2007-09-21 11:59:39";"8.590"
"2007-09-21 12:01:39";"8.480"
"2007-09-21 12:03:39";"8.440"
"2007-09-21 12:05:39";"8.420"
"2007-09-21 12:07:39";"8.370"
"2007-09-21 12:09:39";"8.420"
"2007-09-21 12:11:39";"8.420"
"2007-09-21 12:13:39";"8.640"
"2007-09-21 12:15:39";"8.710"
"2007-09-21 12:17:39";"8.880"
"2007-09-21 12:19:39";"8.840"
"2007-09-21 12:21:39";"8.840"
"2007-09-21 12:23:39";"9.420"
"2007-09-21 12:25:39";"8.850"
"2007-09-21 12:27:39";"8.740"

对于每个任意时间间隔（例如 FROM "2007-09-21 11:15:39" TO "2007-09-21 12:19:39" ）我想提取温度的趋势。

预期结果如下图所示：

在此处输入图像描述

有什么建议吗？

4

1 回答 1

1 
library(MASS)
tem = read.table("temperature.csv",header=TRUE,sep=";",as.is=TRUE)
tem$dt =as.POSIXct(tem$Date_And_Time)
plot(tem$dt,tem$Temperature,type="l")

#Overall trend; use dt
coef(rlm(Temperature~dt,data=tem))

#Trend in first range
coef(rlm(Temperature~dt,data=tem, subset=Date_And_Time < "2007-09-21 10:57:39"))
于 2012-12-31T10:30:22.747 回答