Metadata-Version: 2.4
Name: pychronicles
Version: 0.2.0
Summary: A package for chronicle recognition.
Author-email: Thomas Guyet <thomas.guyet@inria.fr>
Maintainer-email: Thomas Guyet <thomas.guyet@inria.fr>
License:                    GNU LESSER GENERAL PUBLIC LICENSE
                               Version 3, 29 June 2007
        
         Copyright (C) 2007 Free Software Foundation, Inc. <https://fsf.org/>
         Everyone is permitted to copy and distribute verbatim copies
         of this license document, but changing it is not allowed.
        
        
          This version of the GNU Lesser General Public License incorporates
        the terms and conditions of version 3 of the GNU General Public
        License, supplemented by the additional permissions listed below.
        
          0. Additional Definitions.
        
          As used herein, "this License" refers to version 3 of the GNU Lesser
        General Public License, and the "GNU GPL" refers to version 3 of the GNU
        General Public License.
        
          "The Library" refers to a covered work governed by this License,
        other than an Application or a Combined Work as defined below.
        
          An "Application" is any work that makes use of an interface provided
        by the Library, but which is not otherwise based on the Library.
        Defining a subclass of a class defined by the Library is deemed a mode
        of using an interface provided by the Library.
        
          A "Combined Work" is a work produced by combining or linking an
        Application with the Library.  The particular version of the Library
        with which the Combined Work was made is also called the "Linked
        Version".
        
          The "Minimal Corresponding Source" for a Combined Work means the
        Corresponding Source for the Combined Work, excluding any source code
        for portions of the Combined Work that, considered in isolation, are
        based on the Application, and not on the Linked Version.
        
          The "Corresponding Application Code" for a Combined Work means the
        object code and/or source code for the Application, including any data
        and utility programs needed for reproducing the Combined Work from the
        Application, but excluding the System Libraries of the Combined Work.
        
          1. Exception to Section 3 of the GNU GPL.
        
          You may convey a covered work under sections 3 and 4 of this License
        without being bound by section 3 of the GNU GPL.
        
          2. Conveying Modified Versions.
        
          If you modify a copy of the Library, and, in your modifications, a
        facility refers to a function or data to be supplied by an Application
        that uses the facility (other than as an argument passed when the
        facility is invoked), then you may convey a copy of the modified
        version:
        
           a) under this License, provided that you make a good faith effort to
           ensure that, in the event an Application does not supply the
           function or data, the facility still operates, and performs
           whatever part of its purpose remains meaningful, or
        
           b) under the GNU GPL, with none of the additional permissions of
           this License applicable to that copy.
        
          3. Object Code Incorporating Material from Library Header Files.
        
          The object code form of an Application may incorporate material from
        a header file that is part of the Library.  You may convey such object
        code under terms of your choice, provided that, if the incorporated
        material is not limited to numerical parameters, data structure
        layouts and accessors, or small macros, inline functions and templates
        (ten or fewer lines in length), you do both of the following:
        
           a) Give prominent notice with each copy of the object code that the
           Library is used in it and that the Library and its use are
           covered by this License.
        
           b) Accompany the object code with a copy of the GNU GPL and this license
           document.
        
          4. Combined Works.
        
          You may convey a Combined Work under terms of your choice that,
        taken together, effectively do not restrict modification of the
        portions of the Library contained in the Combined Work and reverse
        engineering for debugging such modifications, if you also do each of
        the following:
        
           a) Give prominent notice with each copy of the Combined Work that
           the Library is used in it and that the Library and its use are
           covered by this License.
        
           b) Accompany the Combined Work with a copy of the GNU GPL and this license
           document.
        
           c) For a Combined Work that displays copyright notices during
           execution, include the copyright notice for the Library among
           these notices, as well as a reference directing the user to the
           copies of the GNU GPL and this license document.
        
           d) Do one of the following:
        
               0) Convey the Minimal Corresponding Source under the terms of this
               License, and the Corresponding Application Code in a form
               suitable for, and under terms that permit, the user to
               recombine or relink the Application with a modified version of
               the Linked Version to produce a modified Combined Work, in the
               manner specified by section 6 of the GNU GPL for conveying
               Corresponding Source.
        
               1) Use a suitable shared library mechanism for linking with the
               Library.  A suitable mechanism is one that (a) uses at run time
               a copy of the Library already present on the user's computer
               system, and (b) will operate properly with a modified version
               of the Library that is interface-compatible with the Linked
               Version.
        
           e) Provide Installation Information, but only if you would otherwise
           be required to provide such information under section 6 of the
           GNU GPL, and only to the extent that such information is
           necessary to install and execute a modified version of the
           Combined Work produced by recombining or relinking the
           Application with a modified version of the Linked Version. (If
           you use option 4d0, the Installation Information must accompany
           the Minimal Corresponding Source and Corresponding Application
           Code. If you use option 4d1, you must provide the Installation
           Information in the manner specified by section 6 of the GNU GPL
           for conveying Corresponding Source.)
        
          5. Combined Libraries.
        
          You may place library facilities that are a work based on the
        Library side by side in a single library together with other library
        facilities that are not Applications and are not covered by this
        License, and convey such a combined library under terms of your
        choice, if you do both of the following:
        
           a) Accompany the combined library with a copy of the same work based
           on the Library, uncombined with any other library facilities,
           conveyed under the terms of this License.
        
           b) Give prominent notice with the combined library that part of it
           is a work based on the Library, and explaining where to find the
           accompanying uncombined form of the same work.
        
          6. Revised Versions of the GNU Lesser General Public License.
        
          The Free Software Foundation may publish revised and/or new versions
        of the GNU Lesser General Public License from time to time. Such new
        versions will be similar in spirit to the present version, but may
        differ in detail to address new problems or concerns.
        
          Each version is given a distinguishing version number. If the
        Library as you received it specifies that a certain numbered version
        of the GNU Lesser General Public License "or any later version"
        applies to it, you have the option of following the terms and
        conditions either of that published version or of any later version
        published by the Free Software Foundation. If the Library as you
        received it does not specify a version number of the GNU Lesser
        General Public License, you may choose any version of the GNU Lesser
        General Public License ever published by the Free Software Foundation.
        
          If the Library as you received it specifies that a proxy can decide
        whether future versions of the GNU Lesser General Public License shall
        apply, that proxy's public statement of acceptance of any version is
        permanent authorization for you to choose that version for the
        Library.
        
Project-URL: homepage, https://gitlab.inria.fr/tguyet/pychronicles
Project-URL: documentation, https://tguyet.gitlabpages.inria.fr/pychronicles/
Project-URL: repository, https://gitlab.inria.fr/tguyet/pychronicles
Keywords: data analysis,pattern,timed sequence,temporal data
Classifier: License :: OSI Approved :: GNU Lesser General Public License v3 (LGPLv3)
Classifier: Programming Language :: Python
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.7
Classifier: Programming Language :: Python :: 3.8
Classifier: Programming Language :: Python :: 3.9
Classifier: Programming Language :: Python :: 3.10
Classifier: Operating System :: OS Independent
Requires-Python: >=3.7
Description-Content-Type: text/markdown
License-File: LICENCE.txt
Requires-Dist: pandas>=1.3.0
Requires-Dist: numpy>=1.21.0
Requires-Dist: scipy>=1.7.0
Requires-Dist: lazr.restfulclient>=0.14.5
Requires-Dist: lazr.uri>=1.0.6
Requires-Dist: typing-extensions>=4.4.0
Requires-Dist: metric-temporal-logic>=0.4.0
Requires-Dist: lark>=1.1.4
Requires-Dist: multiset>=3.0.1
Requires-Dist: networkx>=2.6
Requires-Dist: matplotlib>=3.5.0
Provides-Extra: build
Requires-Dist: build; extra == "build"
Requires-Dist: twine; extra == "build"
Provides-Extra: dev
Requires-Dist: pytest; extra == "dev"
Requires-Dist: pytest-cov; extra == "dev"
Dynamic: license-file

---
title: README
author: Thomas Guyet
---

[![PyPI version](https://badge.fury.io/py/pychronicles.svg)](https://badge.fury.io/py/pychronicles)
[![Downloads](https://static.pepy.tech/badge/pychronicles)](https://pepy.tech/project/pychronicles)
[![GitHub license](https://img.shields.io/static/v1?label=Licence&message=LGPL-3&color=green)](https://www.gnu.org/licenses/lgpl-3.0.txt)
[![coverage report](https://gitlab.inria.fr/tguyet/pychronicles/badges/master/coverage.svg)](https://gitlab.inria.fr/tguyet/pychronicles/-/commits/master)
[![Binder](https://mybinder.org/badge_logo.svg)](https://mybinder.org/v2/git/https%3A%2F%2Fgitlab.inria.fr%2Ftguyet%2Fpychronicles/HEAD?labpath=examples%2Fcarepathway_example.ipynb)

# PyChronicles package

A chronicle is a specification of complex temporal behaviors as a graph of temporal constraints. More specifically, a chronicle is a multiset of events and a set of temporal constraints specifying that the occurrences of pairs of events must occur within given temporal intervals.
It can be used to recognize complex behaviors in sequences of temporal events.
PyChronicles is now compatible with Pandas and makes chronicle features [available on dataframes](#dataframe-accessor) representing timed sequences.

This library provides Python classes to define timed sequences and chronicles, and to read/save chronicles in the standard CRS format. The main functionality for chronicles is their efficient matching in timed sequences, or in pandas dataframes (through an accessor).
The `Abstracter` class makes it possible to create a chronicle from a collection of sequences.

For more details about the usage of this package, have a look at the [documentation](https://tguyet.gitlabpages.inria.fr/pychronicles/). For more (formal) information about the temporal model of chronicles, see our [book](https://link.springer.com/book/10.1007/978-3-031-33693-5):
~~~bibtex
@book{chroniclesbook,
  title = {Chronicles: Formalization of a Temporal Model},
  author = {Besnard, Philippe and Guyet, Thomas},
  publisher={Springer Nature},
  pages = {121},
  year = {2023}
}
~~~

## Package installation

The [PyChronicles package](https://pypi.org/project/pychronicles) is available in the [Python Package Index](https://pypi.org/) and can be installed using any standard Python package manager.
~~~sh
pip install pychronicles
~~~

It can also be installed from source. For example, to install it with pip, either locally or in a virtual environment, run the following commands:

~~~sh
git clone --recursive https://gitlab.inria.fr/tguyet/pychronicles
cd pychronicles
pip install -r requirements.txt
pip install --upgrade .
~~~

Alternatively, you can install the repository from sources with pip only: 
~~~sh
git clone --recursive https://gitlab.inria.fr/tguyet/pychronicles
cd pychronicles
pip install .
~~~

In case you want to visualize chronicles, you need to install `networkx` and `matplotlib` manually. By default, they are not installed and the plotting function is disabled.
~~~sh
pip install networkx matplotlib
~~~

## Basic Usage Example

```python
import numpy as np
from pychronicles import TimedSequence, Chronicle

# create a timed sequence
seq = [('a',1),('c',2),('b',3),('a',8)]
ts = TimedSequence(np.array([e[1] for e in seq], dtype='float'), np.array([e[0] for e in seq]))

# create a chronicle
c=Chronicle()
c.add_event(0,'a')
c.add_event(1,'b')
c.add_constraint(0,1, (0.4,2.3))

reco=c.match(ts)
print(f"Does the chronicle match the sequence? [{reco}]")

reco=c.recognize(ts)
print(f"What are the occurrences of the chronicle in the sequence? [{reco}]")
```


## Efficient chronicle recognition

The chronicle recognition package is a pure Python package implemented using numpy features.

### Timed Sequences

A timed sequence is created from a pair of arrays: a numpy array containing the dates, and an array containing the events (with the same length).

The dates can be represented with dates (only with the `np.datetime64` type) or with floats. Integer dates must be converted into floats (the pandas accessor does it automatically).

The type of the events must be `str` or `int`.

_Usage Example_:
```python
>>> seq = [('a',1),('c',2),('b',3),('a',8),('a',10),('b',12),('a',15),('c',17), ('b',20),('c',23),('c',25),('b',26),('c',28),('b',30)]
>>> dates = np.array([np.datetime64('1970-01-01') + np.timedelta64(e[1],'D') for e in seq], dtype='datetime64')
>>> data = np.array([e[0] for e in seq])
>>> ts = TimedSequence(dates, data)
```

### Chronicles

The core of this package is the `Chronicle` class, which represents the chronicle temporal model and offers efficient matching functionalities.


Chronicles handle two models of time (which must be consistent with the model of time of the timed sequences):
* numerical timestamps using floats,
* dates using the `datetime64` format. In this case, the temporal constraints of a chronicle must be defined using `np.timedelta64` values.

A chronicle cannot combine constraints of the two different kinds. The first defined constraint defines the model of time of the chronicle. 

<br>

The package implements efficient algorithms to recognize chronicles. There are three different ways to _recognize_ a chronicle in a sequence of events:

* the absence/presence recognition (`c.match(seq)`): its result is a boolean stating whether the chronicle occurs at least once in the sequence. This is the most efficient algorithm, as it stops as soon as an occurrence is found.
* the occurrence enumeration (`c.recognize(seq)`): its result is the list of the occurrences of the chronicle in the sequence. Contrary to the first function, it looks for all the possible combinations of events. Thus, it is less efficient, but more informative.
* the lazy occurrence enumeration (`c.iter_recognize(seq)`): it returns an iterator over the same occurrences, which are computed on demand. This is useful to process the occurrences one at a time, or to stop the enumeration early.

In an occurrence, two identical events of a chronicle cannot match the same event of the sequence.

Note that a chronicle is somehow similar to a simple temporal network, and the set of constraints may be inconsistent or redundant. It is possible to _minimize_ the temporal constraints of a chronicle using the corresponding function.

```python
>>> c=Chronicle()
>>> c.add_event(0,'a')
>>> c.add_event(1,'b')
>>> c.add_event(2,'c')
>>> c.add_constraint(0,1, (1.0,2.0))
>>> c.add_constraint(0,2, (2.0,5.0))
>>> c.add_constraint(1,2, (0.0,2.0))
>>> print(c)
C1       {{[a],[b],[c]}}
0,1: (1.0, 2.0)
0,2: (2.0, 5.0)
1,2: (0.0, 2.0)

>>> c.minimize()
>>> print(c)
C1       {{[a],[b],[c]}}
0,1: (1.0, 2.0)
0,2: (2.0, 4.0)
1,2: (0.0, 2.0)
```

You can also depict a chronicle as a matplotlib graph:

```python
>>> c.draw()
```


### Fuzzy chronicles

The `FuzzyChronicle` class enables the approximate recognition of a chronicle. A fuzzy chronicle is defined in the same way as a chronicle.
The temporal model is enriched with the modeling of the fuzziness of the temporal constraints (`lbda` parameter).

In addition to the matching functions presented above, it provides a `cmp` function that finds all the occurrences of a chronicle whose degree of matching is above a threshold (parameter `threshold`), together with their similarity. The `iter_cmp` function is its lazy version.

More details about fuzzy chronicles are available in the [article](https://inria.hal.science/hal-02422796v1) (in French):
```bibtex
@inproceedings{fchronicles,
  title = {{\'E}num{\'e}ration des occurrences d'une chronique},
  author = {Guyet, Thomas and Besnard, Philippe and Samet, Ahmed and Ben Salha, Nasreddine and Lachiche, Nicolas},
  booktitle = {Actes de la conférence Extraction et Gestion des Connaissances},
  publisher = {{\'E}ditions RNTI},
  pages = {253--260},
  year = {2020}
}
```

# Dataframe accessor

The `pychronicles` package includes a new accessor for pandas dataframes.
This accessor, denoted `tpattern`, makes it possible to use chronicle functionalities directly with pandas. In this case, you do not even need to manage your timed sequences by hand: everything can be done directly with pandas.

The pandas accessor enables:

* to match or enumerate occurrences of a chronicle
* to match a Metric Temporal Logic (MTL) formula
* to abstract a collection of sequences 
* to extract the delays of matching occurrences of a chronicle


## Spirit of the accessor

The spirit of our accessor is to use a pandas dataframe to encode a sequence or a collection of sequences. The index of the dataframe models the time: it can be an integer, float or date index.

A column of the dataframe can be defined to specify the identifier of the sequences. In the following example, the column named `id` denotes this identifier. Intuitively, it can be used to make `groupby` operations, to process a collection of timed sequences.

The important feature of our package is that any other column can be added to the dataframe. The idea behind this modeling is that the dataframe gathers all the information about the events. The columns describe each event (and can contain `None`).


The example below illustrates the representation of a collection of sequences with a Pandas dataframe.
The basic sequence is repeated three times to create a collection of three timed sequences. In this case, the index of the dataframe is defined with dates.
The event names of the basic sequence feed the column `label`, and we added three columns:


* a column `str_val` with the string values (twice the label)
* a column `num_val` with random numbers
* a column `id` that represents the identifier of a sequence.

```python
# create a basic sequence
seq = [('a',1),('c',2),('b',3),('a',8),('a',10),('b',12),('a',15),('c',17)]
df = pd.DataFrame({
    "label": [e[0] for e in seq]*3,
    "str_val": [e[0]*2 for e in seq]*3,
    "num_val": np.random.randint(10,size=3*len(seq)),
    'id': [1]*len(seq)+[2]*len(seq)+[3]*len(seq)
    },
    index = [np.datetime64('1980-01-01') + np.timedelta64(e[1],'D') for e in seq ]*3
)
```

<br>

Then, the events of a chronicle are no longer event names, but descriptions of the events of interest through their features.
When used with pandas dataframes, the events of a chronicle are queries on the dataframe.
In the example above, an event could be defined by "`label=='a' & num_val>3`".

For the time being, our framework handles string and number attributes. The specification of an event uses the classical operators defined in the [Pandas documentation](https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.query.html). 

## Matching / enumerating the occurrences of a chronicle

Let us now use a dataframe defined as explained in the previous section with a chronicle.

For that, we first need to define a chronicle. It works exactly as before, but the chronicle's events are defined by queries.

For instance, the following example illustrates the definition of a simple chronicle that uses queries as events:
```python
c=Chronicle()
c.add_event(0,'label=="a"')
c.add_event(1,'label=="b" & num_val>5')
c.add_constraint(0,1, (np.timedelta64(4,'D'),np.timedelta64(10,'D')))
``` 

For more details about the syntax of queries, please refer to the [Pandas documentation](https://pandas.pydata.org/docs/reference/api/pandas.DataFrame.query.html). 

This chronicle is semantically sound to be used with the dataframe defined in the example of the previous section. 

### Dataframe representing a sequence

In this section, we consider that the dataframe represents a unique sequence (for instance, we selected the first sequence of the previous dataframe with `df[df.id==1]`).

The recognition functionalities are now accessible through the Pandas dataframe:

* `df.tpattern.match(c)` returns a boolean indicating whether the chronicle occurs in the sequence or not
* `df.tpattern.recognize(c)` returns the list of the occurrences of `c` in the sequence (`df.tpattern.iter_recognize(c)` is its lazy version)
* `df.tpattern.project(c)` returns a dataframe with the delays of the temporal constraints of the chronicle `c` in each of its occurrences

Note that the `tpattern` keyword gives access to the features of our package.

<br>

Using chronicles becomes very easy with Pandas dataframes, and they also gain in expressivity thanks to the use of queries as events.
We keep a good computational efficiency of the recognition algorithms thanks to the rewriting principle developed with semantic chronicles (see the reference below). 

```bibtex
@inproceedings{semantic_chronicles,
  title = {An extension of chronicles temporal model with taxonomies -- Application to epidemiological studies},
  author = {Bakalara, Johanne and Guyet, Thomas and Dameron, Olivier and Happe, Andr{\'e} and Oger, Emmanuel},
  booktitle = {Proceedings of the 14th International Conference on Health Informatics (HEALTHINF)},
  pages = {1--10},
  year = {2021}
}
```

### Dataframe representing a collection of sequences

To process a dataframe containing several sequences, you can use the following idiom based on `apply`:

```python
reco=df.groupby('id').apply(lambda d: d.tpattern.match(c))
```

In this case, the attribute `id` is used to select the events for each sequence, and then the `match` algorithm is applied to this selection.
The result is a pandas series containing a boolean value for each sequence id.

The same applies to the `recognize` function.

## Use of MTL formulas

For situation recognition, there are strong connections between chronicles and metric temporal logics such as MTL or TPTL. See the [article](https://hal.archives-ouvertes.fr/hal-03777471):
~~~bibtex
@inproceedings{chronicle_templogics,
  title = {Logical forms of chronicles},
  author = {Guyet, Thomas and Markey, Nicolas},
  booktitle = {Proceedings of the 29th International Symposium on Temporal Representation and Reasoning (TIME)},
  pages = {1--15},
  year = {2022}
}
~~~

To illustrate that the principle of semantic chronicles can be broader, we extended it to the case of MTL formulas. This means that our framework makes it possible to express MTL formulas using queries as events.


The `tpattern` accessor provides a function `match_mtl` that is dedicated to the recognition of an MTL formula in the timed sequence (only for dataframes indexed by floats or integers). This function returns a boolean value. There is no equivalent to the enumeration of the occurrences of a chronicle.

The syntax of an MTL formula is based on the two operators `F` (eventually) and `G` (globally). These operators can be specified with temporal constraints (in brackets). They combine events defined in the same way as chronicle events.

```python
query = ' F(label=="a" & F[2.9,5]( label=="b" & num_val>5 ))'
df.tpattern.match_mtl(query)
```

This feature is based on the [Python implementation of MTL](https://github.com/mvcisback/py-metric-temporal-logic). 

## Abstraction of a sequence

Finally, we also provide a function that abstracts a collection of sequences into a chronicle.

Contrary to the recognition functions that can handle several columns, the abstraction requires a column that corresponds to the event names. In addition, it requires the name of the attribute that corresponds to the identifier of the sequences.
Then, the function, named `abstract`, takes these two arguments and outputs a single chronicle that occurs in all the sequences.


The following example extracts a chronicle from the dataframe introduced above. Its events are based on the `label` attribute.
```python
chro = df.tpattern.abstract('label', 'id')
```

# Additional features

## Export / import of chronicles

It is possible to specify chronicles using the CRS format (see the functions `load` and `to_crs` of the `pychronicles.utils` module). The following code illustrates the syntax for specifying a chronicle in this format.

```
chronicle C27_sub_0[]()
{
    event(Event_Type1[], t006)
	event(Event_Type1[], t004)
    event(Event_Type2[], t002)
    event(Event_Type3[], t001)

    t004-t006 in [17,25]
    t006-t002 in [-16,-10]
    t002-t001 in [14,29]
    t004-t001 in [27,35]
}
```

## Perspectives

* implementation of chronicle mining algorithms
* graphical chronicle editor
* compile the recognition (e.g. with Cython)
* optimized matching of several chronicles at the same time


If you think that this package can be improved, do not hesitate to contact the author to request additional features. He will do his best to maintain the code and improve it continuously.


# Authorship

* **Author:** Thomas Guyet
* **Institution:** Inria/AIstroSight
* **date:** 10/2026

