| Title: | Discovery of Process Models with the Heuristics Miner |
|---|---|
| Description: | Provides the heuristics miner algorithm for process discovery as proposed by Weijters et al. (2011) <doi:10.1109/CIDM.2011.5949453>. The algorithm builds a causal net from an event log created with the 'bupaR' package. Event logs are a set of ordered sequences of events for which 'bupaR' provides the S3 class eventlog(). The discovered causal nets can be visualised as 'htmlwidgets' and it is possible to annotate them with the occurrence frequency or processing and waiting time of process activities. |
| Authors: | Felix Mannhardt [aut, cre] |
| Maintainer: | Felix Mannhardt <[email protected]> |
| License: | MIT + file LICENSE |
| Version: | 0.2.7 |
| Built: | 2025-02-04 04:41:09 UTC |
| Source: | https://github.com/bupaverse/heuristicsminer |
Converts the object to a Petrinet
as.petrinet(obj)as.petrinet(obj)
obj |
The event log to be used. An object of class |
data(L_heur_1) cn <- causal_net(L_heur_1, threshold = .8) pn <- as.petrinet(cn) petrinetR::render_PN(pn)data(L_heur_1) cn <- causal_net(L_heur_1, threshold = .8) pn <- as.petrinet(cn) petrinetR::render_PN(pn)
Computes the input- and output bindings for use in a causal map. Several heuristics may be used to determine
the activities that are activated or consumed by an event. The Flexible Heuristic Miner (FHM) paper describes
a heuristic that looks ahead (or looks back) until the end of the trace and determines those activities as activated
for which no other cause (activity in a causal dependency) is found. This approach is implemented as type nearest.
causal_bindings(eventlog, dependencies, type = c("nearest"))causal_bindings(eventlog, dependencies, type = c("nearest"))
eventlog |
The bupaR event log. |
dependencies |
A dependency matrix obtained, for example, through |
type |
The heuristic used to determine the bindings. Currently only |
A data frame
causal_bindings(L_heur_1, dependencies = dependency_matrix(L_heur_1))causal_bindings(L_heur_1, dependencies = dependency_matrix(L_heur_1))
Function to create a custom map profile based on some event log attribute.
causal_custom( FUN = mean, attribute, units = "", color_scale = "RdPu", color_edges = "red4", ... )causal_custom( FUN = mean, attribute, units = "", color_scale = "RdPu", color_edges = "red4", ... )
FUN |
A summary function to be called on the process time of a specific activity, e.g. mean, median, min, max |
attribute |
The name of the case attribute to visualize (should be numeric) |
units |
Character to be placed after values (e.g. EUR for monetary euro values) |
color_scale |
Name of color scale to be used for nodes. Defaults to RdPu See |
color_edges |
The color used for edges. Defaults to red4. |
... |
Additional arguments forwarded to FUN |
If used for edges, it will show the attribute values which related to the out-going node of the edge.
causal_net(L_heur_1, type_nodes = causal_custom(attribute = "timestamp"), type_edges = causal_custom(attribute = "timestamp"))causal_net(L_heur_1, type_nodes = causal_custom(attribute = "timestamp"), type_edges = causal_custom(attribute = "timestamp"))
Function to create a frequency profile for a process map.
causal_frequency( value = c("absolute", "relative"), color_scale = "PuBu", color_edges = "dodgerblue4" )causal_frequency( value = c("absolute", "relative"), color_scale = "PuBu", color_edges = "dodgerblue4" )
value |
The type of frequency value to be used: absolute, relative (percentage of activity instances). |
color_scale |
Name of color scale to be used for nodes. Defaults to PuBu. See |
color_edges |
The color used for edges. Defaults to dodgerblue4. |
causal_net(L_heur_1, type = causal_frequency("relative"))causal_net(L_heur_1, type = causal_frequency("relative"))
Creates a Causal net, also known as Heuristics net. This is similar to a processmapR process map.
However, the causal map deals with parallelism by trying to identifying causal dependencies
between activities by using different heuristics as documented in dependency_matrix.
causal_net( eventlog = NULL, dependencies = dependency_matrix(eventlog = eventlog, threshold = threshold, threshold_frequency = threshold_frequency, ...), bindings = causal_bindings(eventlog, dependencies), threshold = 0.9, threshold_frequency = 0, type = causal_frequency("absolute"), sec = NULL, type_nodes = type, type_edges = type, sec_nodes = sec, sec_edges = sec, ... )causal_net( eventlog = NULL, dependencies = dependency_matrix(eventlog = eventlog, threshold = threshold, threshold_frequency = threshold_frequency, ...), bindings = causal_bindings(eventlog, dependencies), threshold = 0.9, threshold_frequency = 0, type = causal_frequency("absolute"), sec = NULL, type_nodes = type, type_edges = type, sec_nodes = sec, sec_edges = sec, ... )
eventlog |
The event log for which a causal map should be computed.
Can be left NULL for more control if parameters |
dependencies |
A dependency matrix created for the event log, for example, by |
bindings |
Causal bindings created by |
threshold |
The dependency threshold to be used when using the default dependency matrix computation. |
threshold_frequency |
The frequency threshold to be used when using the default dependency matrix computation. |
type |
A causal map type. For example, |
sec |
A causal process map type. Values are shown between brackets. |
type_nodes |
A causal map type to be used for nodes only. |
type_edges |
A causal map type to be used for edges only. |
sec_nodes |
A secondary causal map type for nodes only. |
sec_edges |
A secondary causal map type for edges only. |
... |
Further parameters forwarded to the default |
Warning: Projected frequencies are heuristically determined and counts may not add up.
A DiagrammeR graph of the causal map.
# Causal map with default parameters causal_net(L_heur_1) # Causal map with lower dependency treshold causal_net(L_heur_1, threshold = .8) # For even more control omit the `eventlog` parameter # and provide `dependencies` and `bindings` directly. d <- dependency_matrix(L_heur_1, threshold = .8) causal_net(dependencies = d, bindings = causal_bindings(L_heur_1, d, "nearest")) # The returned DiagrammeR object can be further augmented with # panning and zooming before rendering: library(magrittr) causal_net(L_heur_1) %>% render_causal_net(render = TRUE) %>% DiagrammeRsvg::export_svg() %>% svgPanZoom::svgPanZoom()# Causal map with default parameters causal_net(L_heur_1) # Causal map with lower dependency treshold causal_net(L_heur_1, threshold = .8) # For even more control omit the `eventlog` parameter # and provide `dependencies` and `bindings` directly. d <- dependency_matrix(L_heur_1, threshold = .8) causal_net(dependencies = d, bindings = causal_bindings(L_heur_1, d, "nearest")) # The returned DiagrammeR object can be further augmented with # panning and zooming before rendering: library(magrittr) causal_net(L_heur_1) %>% render_causal_net(render = TRUE) %>% DiagrammeRsvg::export_svg() %>% svgPanZoom::svgPanZoom()
Function to create a performance profile for a causal map.
causal_performance( FUN = mean, units = c("mins", "secs", "hours", "days", "weeks", "months", "quarters", "semesters", "years"), color_scale = "Reds", color_edges = "red4", ... )causal_performance( FUN = mean, units = c("mins", "secs", "hours", "days", "weeks", "months", "quarters", "semesters", "years"), color_scale = "Reds", color_edges = "red4", ... )
FUN |
A summary function to be called on the process time of a specific activity, e.g. mean, median, min, max |
units |
The time unit in which processing time should be presented (mins, hours, days, weeks, months, quarters, semesters, years. A month is defined as 30 days. A quarter is 13 weeks. A semester is 26 weeks and a year is 365 days |
color_scale |
Name of color scale to be used for nodes. Defaults to Reds. See |
color_edges |
The color used for edges. Defaults to red4. |
... |
Additional arguments forwarded to FUN |
causal_net(L_heur_1, type = causal_performance())causal_net(L_heur_1, type = causal_performance())
Creates a dependency matrix from a precedence matrix (precedence_matrix) based on different approaches.
dependency_matrix( eventlog = NULL, dependency_type = dependency_type_fhm(threshold_dependency = threshold, threshold_frequency = threshold_frequency, ...), threshold = 0.9, threshold_frequency = 0, ... )dependency_matrix( eventlog = NULL, dependency_type = dependency_type_fhm(threshold_dependency = threshold, threshold_frequency = threshold_frequency, ...), threshold = 0.9, threshold_frequency = 0, ... )
eventlog |
A bupaR event log, may be NULL when a precedence matrix is provided. |
dependency_type |
Which approach to use for calculation of the dependency matrix. Currently only ( |
threshold |
A dependency threshold, usually in the interval |
threshold_frequency |
An absolute frequency threshold filtering dependencies which are observed infrequently. |
... |
Parameters forwarded to ( |
A square matrix with class dependency_matrix containing the computed dependency values between all activities.
d <- dependency_matrix(L_heur_1) print(d) as.matrix(d)d <- dependency_matrix(L_heur_1) print(d) as.matrix(d)
Computes the dependencies based on the approach known as Flexible Heuristics Miner.
dependency_type_fhm( threshold_dependency = 0.9, threshold_l1 = threshold_dependency, threshold_l2 = threshold_dependency, threshold_frequency = 0, all_connected = FALSE, endpoints_connected = FALSE )dependency_type_fhm( threshold_dependency = 0.9, threshold_l1 = threshold_dependency, threshold_l2 = threshold_dependency, threshold_frequency = 0, all_connected = FALSE, endpoints_connected = FALSE )
threshold_dependency |
A dependency threshold, usually in the interval |
threshold_l1 |
A dependency threshold, usually in the interval |
threshold_l2 |
A dependency threshold, usually in the interval |
threshold_frequency |
An absolute frequency threshold filtering dependencies which are observed infrequently. |
all_connected |
If |
endpoints_connected |
If |
A dependency type.
A. J. M. M. Weijters and J. T. S. Ribeiro, "Flexible Heuristics Miner (FHM)," 2011 IEEE Symposium on Computational Intelligence and Data Mining (CIDM), Paris, 2011, pp. 310-317. doi: 10.1109/CIDM.2011.5949453
dependency_matrix(L_heur_1, dependency_type = dependency_type_fhm(all_connected = TRUE))dependency_matrix(L_heur_1, dependency_type = dependency_type_fhm(all_connected = TRUE))
Computes the dependencies based on the approach taking into account activity durations based on life-cycle transitions.
dependency_type_lifecycle( threshold_dependency = 0.9, threshold_l1 = threshold_dependency, threshold_frequency = 0, all_connected = FALSE, endpoints_connected = FALSE )dependency_type_lifecycle( threshold_dependency = 0.9, threshold_l1 = threshold_dependency, threshold_frequency = 0, all_connected = FALSE, endpoints_connected = FALSE )
threshold_dependency |
A dependency threshold, usually in the interval |
threshold_l1 |
A dependency threshold, usually in the interval |
threshold_frequency |
An absolute frequency threshold filtering dependencies which are observed infrequently. |
all_connected |
If |
endpoints_connected |
If |
A dependency type.
A. Burattin and A. Sperduti, “Heuristics Miner for Time Intervals,” in ESANN 2010, 18th European Symposium on Artificial Neural Networks, Bruges, Belgium, April 28-30, 2010, Proceedings, 2010.
dependency_matrix(L_heur_1, dependency_type = dependency_type_fhm(all_connected = TRUE))dependency_matrix(L_heur_1, dependency_type = dependency_type_fhm(all_connected = TRUE))
Sample of 10 000 traces from an artificial eventlog from the PhD thesis 'Multi-perspective Process Mining' used to illustrate the Data-aware Heuristic Miner algorithm.
hospital_multi_perspectivehospital_multi_perspective
Eventlog containing a sample of 10 000 cases
doi:10.4121/uuid:32cad43f-8bb9-46af-8333-48aae2bea037
Mannhardt, F. (Felix) (2016) Data-driven Process Discovery - Artificial Event Log. Eindhoven University of Technology. Dataset. https://doi.org/10.4121/uuid:32cad43f-8bb9-46af-8333-48aae2bea037
Artificial eventlog for illustrating Heuristics Miner published as supplementary material to the book
Process Mining: Discovery, Conformance and Enhancement of Business Processes.
L_heur_1L_heur_1
Eventlog containing 40 cases
Process Mining: Discovery, Conformance and Enhancement of Business Processes by W.M.P. van der Aalst, Springer Verlag, 2011 (ISBN 978-3-642-19344-6).
Artificial eventlog for illustrating Heuristics Miner published as supplementary material to the book
Process Mining: Discovery, Conformance and Enhancement of Business Processes.
L_heur_2L_heur_2
Eventlog containing 85 cases
Process Mining: Discovery, Conformance and Enhancement of Business Processes by W.M.P. van der Aalst, Springer Verlag, 2011 (ISBN 978-3-642-19344-6).
Parallel Matrix with Lifecycle
parallel_matrix_lifecycle(eventlog)parallel_matrix_lifecycle(eventlog)
eventlog |
The event log object to be used. |
parallel_matrix_lifecycle(L_heur_1)parallel_matrix_lifecycle(L_heur_1)
Visualize a dependency matrix. A generic plot function for dependency matrices.
## S3 method for class 'dependency_matrix' plot(x, ...)## S3 method for class 'dependency_matrix' plot(x, ...)
x |
Dependency matrix |
... |
Additional parameters |
A ggplot object, which can be customized further, if deemed necessary.
Construct a precedence matrix, showing how activities are followed by each other.
This is a performance improved variant of precedence_matrix
in the processmapR package.
precedence_matrix( eventlog, type = c("absolute", "relative", "relative-antecedent", "relative-consequent", "relative-case") )precedence_matrix( eventlog, type = c("absolute", "relative", "relative-antecedent", "relative-consequent", "relative-case") )
eventlog |
The event log object to be used |
type |
The type of precedence matrix, which can be absolulte, relative, relative-antecedent or relative-consequent. Absolute will return a matrix with absolute frequencies, relative will return global relative frequencies for all antecedent-consequent pairs. Relative-antecedent will return relative frequencies within each antecendent, i.e. showing the relative proportion of consequents within each antecedent. Relative-consequent will do the reverse. |
m <- precedence_matrix(hospital_multi_perspective, type = "absolute") print(m) as.matrix(m)m <- precedence_matrix(hospital_multi_perspective, type = "absolute") print(m) as.matrix(m)
Construct a precedence matrix, showing how activities are followed by each other.
This function computes the precedence matrix directly in C++ for efficiency.
Only the type absolute of (precedence_matrix) is supported.
precedence_matrix_absolute(eventlog, lead = 1)precedence_matrix_absolute(eventlog, lead = 1)
eventlog |
The event log object to be used. |
lead |
The distance between activities following/preceding each other. |
library(eventdataR) data(traffic_fines) m <- precedence_matrix_absolute(traffic_fines) print(m) as.matrix(m)library(eventdataR) data(traffic_fines) m <- precedence_matrix_absolute(traffic_fines) print(m) as.matrix(m)
Construct a precedence matrix counting how often pattern aba occurs.
precedence_matrix_length_two_loops(eventlog)precedence_matrix_length_two_loops(eventlog)
eventlog |
The event log object to be used. |
m <- precedence_matrix_length_two_loops(hospital_multi_perspective) print(m) as.matrix(m)m <- precedence_matrix_length_two_loops(hospital_multi_perspective) print(m) as.matrix(m)
Precedence Matrix with Lifecycle
precedence_matrix_lifecycle(eventlog)precedence_matrix_lifecycle(eventlog)
eventlog |
The event log object to be used. |
precedence_matrix_lifecycle(L_heur_1)precedence_matrix_lifecycle(L_heur_1)
Generic print function for a Causal net
## S3 method for class 'causal_net' print(x, ...)## S3 method for class 'causal_net' print(x, ...)
x |
Causal net object |
... |
Additional Arguments |
Generic print function for a dependency matrix
## S3 method for class 'dependency_matrix' print(x, ...)## S3 method for class 'dependency_matrix' print(x, ...)
x |
dependency matrix object |
... |
Additional Arguments |
Renders a Causal net as graph
render_causal_net( causal_net, rankdir = "LR", layout = "dot", render = T, fixed_edge_width = F, fixed_node_pos = NULL, ... )render_causal_net( causal_net, rankdir = "LR", layout = "dot", render = T, fixed_edge_width = F, fixed_node_pos = NULL, ... )
causal_net |
A causal net created by |
rankdir |
Rankdir to be used for DiagrammeR. |
layout |
Layout to be used for DiagrammeR. |
render |
Whether to directly render the DiagrammeR graph or simply return it. |
fixed_edge_width |
If TRUE, don't vary the width of edges. |
fixed_node_pos |
When specified as a data.frame with three columns 'act', 'x', and 'y' the position of nodes is fixed. Note that his can only be used with the 'neato' layout engine. |
... |
Further parameters forwarded to the DiagrammeR render function. |
A DiagrammeR graph of the Causal net.
render_causal_net(causal_net(L_heur_1))render_causal_net(causal_net(L_heur_1))
Creates a dependency graph visualizing the contents of a dependency matrix.
render_dependency_matrix( dependencies, rankdir = "LR", layout = "dot", render = T )render_dependency_matrix( dependencies, rankdir = "LR", layout = "dot", render = T )
dependencies |
A dependency matrix created by |
rankdir |
Rankdir to be used for DiagrammeR. |
layout |
Layout to be used for DiagrammeR. |
render |
Whether to directly render the DiagrammeR graph or simply return it. |
A DiagrammeR graph of the (filtered) dependency matrix.
render_dependency_matrix(dependency_matrix(L_heur_1))render_dependency_matrix(dependency_matrix(L_heur_1))