Model clusters and their temporal relationships using an extensible Markov
chain. See vignette("rEMM") for a getting started guide.
Models
EMM combines data stream clustering with a temporal model.
tNN provides threshold nearest neighbor clustering, and
TRACDS learns temporal relationships from cluster assignments.
Temporal Structure for Standard Clustering Algorithms
TRAC adds a Markov model to an existing clustering of
sequence data, such as a k-means or PAM clustering.
Building models
build adds observations to an EMM,
cluster adds observations to a tNN clustering, and
update adds cluster assignments to a temporal model.
Working with the models
plot visualizes a model, and find_clusters
assigns observations to existing clusters without updating the model.
recluster groups micro-clusters, and
merge_clusters merges specified states.
remove removes specified states or transitions,
prune removes rare states or transitions, and
fade reduces the weight of older observations.
combine combines EMMs.
Transitions and sequence analysis
transition accesses transition probabilities and counts,
and transition_table extracts transitions for a new sequence.
smooth_transitions smooths transition counts between
neighboring states. score measures how well a sequence agrees
with a model, and predict predicts future states.
Data
synthetic_stream generates synthetic sequence data.
Bundled examples include EMMsim (simulated sequences),
EMMTraffic (hypothetical traffic observations),
Derwent (river flow readings), and
rRNA16S (16S rRNA sequence counts).
Author
Maintainer: Michael Hahsler mhahsler@lyle.smu.edu (ORCID) [copyright holder]
Authors:
Michael Hahsler mhahsler@lyle.smu.edu (ORCID) [copyright holder]
Other contributors:
Margaret H. Dunham [contributor]