stream (version 1.3-2)

DSC_EA: Evolutionary Algorithm

Description

Reclustering using an evolutionary algorithm. This approach was designed for evoStream but can also be used for other micro-clustering algorithms. The evolutionary algorithm uses existing clustering solutions and creates small variations of them by combining and randomly modfiying them. The modified solutions can yield better partitions and thus can improve the clustering over time. The evolutionary algorithm is incremental, which allows to improve existing macro-clusters instead of recomputing them every time.

Usage

DSC_EA(k, generations = 2000, crossoverRate = 0.8,
  mutationRate = 0.001, populationSize = 100)

Arguments

k

number of macro-clusters

generations

number of EA generations performed during reclustering

crossoverRate

cross-over rate for the evolutionary algorithm

mutationRate

mutation rate for the evolutionary algorithm

populationSize

number of solutions that the evolutionary algorithm maintains

References

Carnein M. and Trautmann H. (2018), "evoStream - Evolutionary Stream Clustering Utilizing Idle Times", Big Data Research.

Examples

# NOT RUN {
stream <- DSD_Memory(DSD_Gaussians(k = 3, d = 2), 1000)

## online algorithm
dbstream <- DSC_DBSTREAM(r=0.1)

## offline algorithm
EA <- DSC_EA(k=3, generations=1000)

## create pipeline and insert observations
two <- DSC_TwoStage(dbstream, EA)
update(two, stream, n=1000)

## plot resut
reset_stream(stream)
plot(two, stream, type="both")

## if we have time, evaluate additional generations. This can be
## called at any time, also between observations.
two$macro_dsc$RObj$recluster(2000)

## plot improved result
reset_stream(stream)
plot(two, stream, type="both")


## alternatively: do not create twostage but apply directly
reset_stream(stream)
update(dbstream, stream, n = 1000)
recluster(EA, dbstream)
reset_stream(stream)
plot(EA, stream)


# }