Hi I am forecasting data based on interpolated data. My interpolated data: metervalue of dataID=1103data as below:
dataid,meter_value
1103,183138
1103,183138
1103,183138
1103,183138
:
:
1103,197015.9857
1103,197015.9928
1103,197016
output of time(zz) as below.
x,dataid,meter_value
27/10/2015 19:50
27/10/2015 19:51
27/10/2015 19:52
27/10/2015 19:53
:
:
24/2/2016 0:38
24/2/2016 0:39
24/2/2016 0:40
Based on that interpolated data,I am trying to forecast 6 months data starting from 24/02/2016 0:41 to 24/07/2016 0:41.
I tried to create myts with 1min frequency interval for predicting data using ETS and HoltWinter but, all forecasting plot are weird for 1 min interval data. I could not get 80% and 95% high low data range with predicted data.
My interploated data as below:
to.minute <- function(x) as.POSIXct(trunc(as.POSIXct(x, origin = "1970-01-01"), "mins"))
z <- read.csv.zoo("test_1103.csv", FUN = to.minute, aggregate = function(x) tail(x, 1))
zz <- na.approx(as.zoo(as.ts(z)))
time(zz) <- as.POSIXct(time(zz), origin = "1970-01-01")
My forecasting data as below:
myts <- ts(zz$meter_value, start = c(2015,10,27,19,50), end = c(2016,02,24,0,40), frequency = 10080)
forecast(myts,12)
Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
2016.00020 183248.6 183248.6 183248.6 183248.6 183248.6
2016.00030 183248.6 183248.6 183248.6 183248.6 183248.6
2016.00040 183248.6 183248.6 183248.6 183248.6 183248.6
2016.00050 183248.6 183248.6 183248.6 183248.6 183248.7
2016.00060 183248.6 183248.6 183248.7 183248.6 183248.7
2016.00069 183248.6 183248.6 183248.7 183248.6 183248.7
2016.00079 183248.6 183248.6 183248.7 183248.6 183248.7
2016.00089 183248.7 183248.6 183248.7 183248.6 183248.7
2016.00099 183248.7 183248.6 183248.7 183248.6 183248.7
2016.00109 183248.7 183248.6 183248.7 183248.6 183248.8
2016.00119 183248.7 183248.6 183248.7 183248.6 183248.8
2016.00129 183248.7 183248.6 183248.8 183248.6 183248.8
How could i change the format of point forecast to "yyyy-mm-dd hh:mm" what is the meaning of "four decimal values" (eg.2016.00129) in point forecast ?
Same for HoltWinter Forecast method: I could not get the plot with 80% and 95% range interval ad prediction data.
HoltWinters(x = myts, beta = TRUE, gamma = FALSE)
Smoothing parameters:
alpha: 0.9999424
beta : TRUE
gamma: FALSE
Coefficients:
[,1]
a 1.832486e+05
b 6.906077e-03
forecast(hw, 12)
Point Forecast Lo 80 Hi 80 Lo 95 Hi 95
2016.00020 183248.6 183248.6 183248.6 183248.6 183248.6
2016.00030 183248.6 183248.6 183248.6 183248.6 183248.6
2016.00040 183248.6 183248.6 183248.6 183248.6 183248.6
2016.00050 183248.6 183248.6 183248.6 183248.6 183248.7
2016.00060 183248.6 183248.6 183248.7 183248.6 183248.7
2016.00069 183248.6 183248.6 183248.7 183248.6 183248.7
2016.00079 183248.6 183248.6 183248.7 183248.6 183248.7
2016.00089 183248.7 183248.6 183248.7 183248.6 183248.7
2016.00099 183248.7 183248.6 183248.7 183248.6 183248.7
2016.00109 183248.7 183248.6 183248.7 183248.6 183248.8
2016.00119 183248.7 183248.6 183248.7 183248.6 183248.8
2016.00129 183248.7 183248.6 183248.8 183248.6 183248.8
Both ETS and HoltWinter Expected forecasted dataplot
Is "myts" with 1 min frequency interval correct? How should i change the plot so that I can see the 80% and 95% range with prediction data.