Implements a sliding window data stream operator which keeps a fixed amount (window length) of the most recent data points of the stream.
Value
An object of class DSAggregate_Window (subclass of DSAggregate).
Details
If lambda is greater than 0 then the weight uses a damped window
model (Zhu and Shasha, 2002). The weight for points in the window follows
\(2^(-lambda*t)\) where \(t\) is the age of the point.
References
Zhu, Y. and Shasha, D. (2002). StatStream: Statistical Monitoring of Thousands of Data Streams in Real Time, Intl. Conference of Very Large Data Bases (VLDB'02).
See also
Other DSAggregate:
DSAggregate(),
DSAggregate_Sample()
Examples
set.seed(1500)
## Example 1: Basic use
stream <- DSD_Gaussians(k = 3, d = 2, noise = 0.05)
window <- DSAggregate_Window(horizon = 10)
window
#> Sliding windowClass: DSAggregate_Window, DSAggregate, DST
# update with only two points. The window is mostly empty (NA)
update(window, stream, 2)
get_points(window)
#> X1 X2 .class
#> 3 NA NA NA
#> 4 NA NA NA
#> 5 NA NA NA
#> 6 NA NA NA
#> 7 NA NA NA
#> 8 NA NA NA
#> 9 NA NA NA
#> 10 NA NA NA
#> 1 0.8762547 0.5229369 2
#> 2 0.8124275 0.2624523 1
# get weights and window as a single data.frame
get_model(window)
#> weight X1 X2 .class
#> 3 1 NA NA NA
#> 4 1 NA NA NA
#> 5 1 NA NA NA
#> 6 1 NA NA NA
#> 7 1 NA NA NA
#> 8 1 NA NA NA
#> 9 1 NA NA NA
#> 10 1 NA NA NA
#> 1 1 0.8762547 0.5229369 2
#> 2 1 0.8124275 0.2624523 1
# update window
update(window, stream, 100)
get_points(window)
#> X1 X2 .class
#> 3 0.4261088 0.2965189 3
#> 4 0.7784012 0.2364471 1
#> 5 0.9029506 0.5557783 2
#> 6 0.7720509 0.2656545 1
#> 7 0.8536844 0.6830931 NA
#> 8 0.9544277 0.4679511 2
#> 9 0.8106137 0.2341865 1
#> 10 0.4306247 0.1629674 3
#> 1 0.8833969 0.5530094 2
#> 2 0.7468827 0.2084139 1
## Example 2: Implement a classifier over a sliding window
window <- DSAggregate_Window(horizon = 100)
update(window, stream, 1000)
# train the classifier on the window
library(rpart)
tree <- rpart(`.class` ~ ., data = get_points(window))
tree
#> n=95 (5 observations deleted due to missingness)
#>
#> node), split, n, deviance, yval
#> * denotes terminal node
#>
#> 1) root 95 58.98947 1.989474
#> 2) X1>=0.6217785 66 16.36364 1.545455
#> 4) X2< 0.3527057 30 0.00000 1.000000 *
#> 5) X2>=0.3527057 36 0.00000 2.000000 *
#> 3) X1< 0.6217785 29 0.00000 3.000000 *
# predict the class for new points from the stream
new_points <- get_points(stream, n = 100, info = FALSE)
pred <- predict(tree, new_points)
plot(new_points, col = pred)
