=Paper=
{{Paper
|id=Vol-2896/RELATED_2021_paper_5
|storemode=property
|title=Characteristics of Effective Paths in Brazilian Legal Processes
|pdfUrl=https://ceur-ws.org/Vol-2896/RELATED_2021_paper_5.pdf
|volume=Vol-2896
|authors=Jorge Luiz Bezerra de Araújo,João A. Monteiro Neto,Fernando Siqueira,Claúdio Moura Santos,Ricardo Vasconcelos,Erneson A. Oliveira,Carlos de Oliveira Caminha Neto,Vasco Furtado
}}
==Characteristics of Effective Paths in Brazilian Legal Processes==
https://ceur-ws.org/Vol-2896/RELATED_2021_paper_5.pdf
Characteristics of Effective Paths in Brazilian Legal
Processes
Jorge Luiz Bezerra de Araújo1 , João A. Monteiro Neto4 , Fernando Siqueira1 ,
Claúdio Moura Santos1 , Ricardo Vasconcelos1 , Erneson A. Oliveira2,31 ,
Carlos de Oliveira Caminha Neto2 and Vasco Furtado2
1
Laboratório de Ciência de Dados e Inteligência Artificial, Universidade de Fortaleza, 60811-905 Fortaleza, Ceará, Brasil.
2
Programa de Pós Graduação em Informática Aplicada, Universidade de Fortaleza, 60811-905 Fortaleza, Ceará, Brasil.
3
Mestrado Profissional em Ciências da Cidade, Universidade de Fortaleza, 60811-905, Fortaleza, Ceará, Brasil.
4
Curso de Direito, Universidade de Fortaleza, 60811-905 Fortaleza, Ceará, Brasil.
Abstract
A Brazilian judicial process is essentially constituted by a sequence of actions that mark the history of
the development of activities within the process. In this article, we create and analyse the most probable
paths that characterize the classes of legal proceedings through a set of rules attributed to a complex
network. We analyze the sequence of activities most probable and highlight the average times required
to complete. We observed what would be the main points of delays in the possible ways of starting
and ending a process in addition to the similarity between the possible paths taken. In addition, we
indicate how much time would be spent, on average, in increasing the number of movements for each
class of process. The characterization of these networks and the contextualization of the ideal movement
path, associated with the other data generated by the proposed method, allow, through a reliable data
generation process, a detailed observation of the flow of lawsuits, the identification of atypical behaviors
and the planning of interventions capable of optimizing the provision of the judicial service by the
judiciary.
Keywords
Complex networks„ Juridic processes, Shortest path
1. Introduction
A judicial proceeding aims to eliminate divergent conflicts of interest by making fair use of the
laws applicable to the matters discussed in the course of the procedural relationship. In general,
a judicial process encompasses a series of steps, called procedural movements, delimited and
guided by the legal and time frames of the process, and are carried out by all the actors involved
in the procedural relationship, Judge, Prosecutor, Lawyers, Parties and Servants of the Judiciary.
Logically structured, but not linear, these movements guide the development of the process
from its beginning to its conclusion. Steps such as the Petição Inicial (Initial Petition), Citação
(Citation), Réplica (Replica), Senteça (Sentence) and Apelação (Appeals) are common movements
RELATED - Relations in the Legal Domain Workshop, in conjunction with ICAIL 2021, June 25, 2021, São Paulo, Brazil
Envelope-Open jlbdearaujo@gmail.com (J. L. B. d. Araújo); joaoneto@unifor.br (J. A. M. Neto); fsiqueira@edu.unifor.br
(F. Siqueira); clausmoura@edu.unifor.br (C. M. Santos); ricardo.vasconcelos@tjce.jus.br (R. Vasconcelos);
erneson@unifor.br (E. A. Oliveira2,3 ); caminha@unifor.br (C. d. O. C. Neto); vasco@unifor.br (V. Furtado)
© 2021 Copyright for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR
Workshop
Proceedings
http://ceur-ws.org
ISSN 1613-0073
CEUR Workshop Proceedings (CEUR-WS.org)
in most processes. In essence, they lead information, in sequential form, such as identification of
who brought the action, verification of the required laws, arguments of the defendant’s defense,
presentation of evidence and the final decision. Such movements depend and vary significantly
depending on the matters involved (criminal law, family law, etc.) which makes it even more
complex to obtain sequential standardization between legal processes. Despite the existence
of a basic standardization taking into account the type of case, the set of transactions is not
influenced exclusively of the judicial classes to which the cases are initially assigned, being often
influenced by the ”history” of the case and the needs that the Judge can having to realize the
most diverse acts that can substantiate their final decision, which thus creates new movements
and cycles of actions necessary for it to be finalized, which are different from the path initially
envisaged.
Typically, the increase in the number of movements within a process makes it take more time
to complete. Thus, the visualization of an optimal or sub-optimal ”procedural path” becomes of
great interest both for the parties involved in the case and particularly for the Judiciary, which
can assess the level of efficiency of its units by identifying possible deviations and designing
interventions to correct identified inconsistencies [1]. Even though these paths may be too
complicated to be observed mathematically, in addition to the complexities of subjective actions
taken by the parties and the judge in the course of a lawsuit, observing which movements
would cause the greatest delays or greatest accelerations can serve as an important element
encouraging for specific actions to maximize the services provided by the judiciary.
Thus, due to the difficulty of establishing a standard sequence of procedural movements, an
alternative methodology for characterizing processes through interactions in complex networks
is presented here [2, 3, 4]. In the section (2) an overview of the proposed problem will be
presented. In the section (3), the mathematical rules for creating complex networks for the
different chosen procedural classes will be presented. We also present the most probable paths
based on data from the Court of Justice (TJ) Procedural Tables. The times spent for each stage
and the connections between the most important paths are highlighted during the text. Finally,
we condensed the results in the section (4) where we highlight an efficient way to obtain the
influence of possible increases in time due to the need for new movements. In the section (5)
we highlight the main points of our research with the main perspectives.
2. State of the art
The theory of complex networks is a highly interdisciplinary area that offers resources for the
study of the most varied types of complex systems. Many dynamic nonlinear processes and
stochastic simulations can be analyzed in complex networks, such as: interactions between
proteins, human relationships, air transport systems, the Internet, the financial market, among
other complex systems [5].
Complex network analysis is a critical tool for understanding various complex systems [6],
including as an approach to obtain a quantitative understanding of the structure and evolution
of law [7]. Computer science scientists and legal experts have used citation analysis methods in
order to build networks of case law citations, as well as to model and quantify the complexity
of the legislative [8].
It has also been used to analyze data from complex networks built from decisions of national
and international courts, statutes, constitutions and international treaties. There are also works
in this context that explore, for example, what characteristics of complex systems occur in
statutory law, how references to judicial decisions are used to shape legal arguments, or where
social dynamics exist between international judges or arbitrators. Chandler [9] examined the
complex network structure of precedent-based court rulings, using data from the United States
Supreme Court.
Although there are countless studies that characterize and analyze data in complex networks,
no reference was found to study the movements of legal processes, in an attempt to identify
and present the most probable paths of legal proceedings through a set of rules attributed to
a complex network. Therefore, one of the contributions of this work is to start from the data
of the classes and movements of the judicial processes of the Ceará state justice, to represent
them as a complex network, modeling the existing relationships in such a way that it becomes
possible to characterize the judicial processes.
3. Model
Data used were based on the classes and movements of the Unified Procedural Tables of the
Judiciary created based on CNJ Resolution No. 46/2007, as well as on the record of internal
movements operated within the scope of Ceará State Justice. The 1st degree lawsuits filed until
2019 and written off in 2019 in the various instances of the Tribunal de Justiça (TJ - Court of
Justice) of the State of Ceará (Brazil) were considered, in the classes: Procedimento Comum Cível
(Common Civil Procedure), Execução Fiscal (Tax Enforcement), Procedimento do Juizado Especial
Cível (Procedure of the Special Civil Court) and Execução de Título Extrajudicial (Execution of
Extrajudicial Title). In total, the number of 4797 cases were analyzed for the Class of Execution
of Extrajudicial Title, 67499 for Common Civil Procedure, 29643 Procedure of the Special Civil
Court and 23039 for Tax Enforcement with at least 300 types of different movements.
Our database from the TJ contains multiple information that characterises the processes. We
emphasize that for this work, data from the judicial classes named as Enforcement of Title,
Common Civil Procedure, Procedure of the Special Civil Court and Tax Enforcement were
used as the most demanded classes of the assessed State Justice. The temporal information for
each movement presented reveals its initializations. Thus, when a new movement starts, it
indicates a new series of judicial activities. Such activities depend on the type of movement.
Furthermore, the emergence of a new movement indicates the end of another, suggesting that
the term ’movement’ is associated with a temporal state of the legal process. In this case, we
label as Δ𝑇 the lasting times of each move within a process. We identify at each step of the
process which is the pair OLD → NOW as the construction of a junction between the type of
movement in a previous sequence (OLD) and the sequence current (NOW ). This, it is possible
to set up a data table containing all flows between types of consecutive movements, in addition
to obtaining data on the times spent between such movements.
To reveal the importance of transitions between different types of movements, we define a
matrix F where its elements 𝐹𝑖𝑗 represent the number of observed strings that initially emerged
from the i step and are in the j step. The matrix F is not symmetric since it is possible to
obtain an element with 𝐹𝑖𝑗 ≠ 𝐹𝑗𝑖 . Thus, the asymmetry of the problem suggests the spatial
construction of a network of targeted actions where the nodes lead the information for each
type of movement and the edges link to the sequential evolutions tuned by 𝐹𝑖𝑗 . In the Figure (1)
an illustration of the formation of the creation of the complex procedural network is presented.
The dots indicate each type of movement and the arrows indicate the transformation vectors
between sequential movements.
FE C C A
FA B
F
Most Probable Paths
FC D B
E D FB E
FB E E
FE C
B A C
FC D
FA B D
Figure 1: On the left side, an imaginary schematic model of a complex directional network for types
of movement between processes is presented. Each node in the network contains a type of process
(illustrated by A, B, C, D, E and F). The different widths of the arrows represent the associated weights
of that network which are regulated by the 𝐹𝑖𝑗 indexes. On the right, a sequence 𝐴 → 𝐵 → 𝐸 → 𝐶 → 𝐷
is displayed indicating the most probable path between moves that start at 𝐴 and end at 𝐷.
Thus, it is intended, for each type of judicial class, to build paths between the movements
that represent it as being built, in essence, by the movements that will be visited.
An indicated way to create the most probable path is to check what would be the most
frequent type of movement for the first stage of each judicial class. From there, the next points
would be accessed by the incidence of 𝐹 thus creating an optimal local path. This type of strategy
is seen in Greedy algorithm techniques [10]. However, in systems with high ramifications and
decision making with probabilities between similar movements, this approach can induce a
large number of sequences of high degeneration or produce suboptimal temporal paths. An
alternative is to check among all possible combinations of paths which would be the one with
the lowest cost, or the most probable to happen. For this purpose, we created a matrix 𝐷 𝑒𝑓 ,
called effective distance, to which its elements are regulated by:
𝑒𝑓
𝐷𝑖𝑗 = 𝑒𝑥𝑝(−𝛽𝐹𝑖𝑗 ) (1)
where 𝛽 = ⟨𝐹1 ⟩ regulates the influence of 𝐹𝑖𝑗 . Thus, we say that the most probable path (PP) of a
𝑖𝑗
𝑒𝑓
process is found minimizing all routes 𝐷𝑖𝑗 , that is:
𝑒𝑓
𝑃𝑃𝑖𝑗 = 𝐷𝑖𝑗𝑘𝑠𝑡𝑟𝑎(𝐷𝑖𝑗 ) (2)
where 𝐷𝑖𝑗𝑘𝑠𝑡𝑟𝑎 [11] is the function that returns the points of the networks with the lowest cost
𝑒𝑓
regulated by 𝐷𝑖𝑗 .
Briefly, we calculate all the frequencies of pairs between consecutive movements and store
them in the matrix 𝐹. Once all the elements of 𝐹 are calculated, it is possible to obtain all the
elements of 𝐷𝑖𝑗 through equation (1). The 𝐷𝑖𝑗𝑘𝑠𝑡𝑟𝑎 algorithm returns all points interspersed in
the path between 𝑖 and 𝑗 with the least cost in the sum of all possible paths. Thus, the most
influential movements for each procedural class will be discarded, thus being able to obtain
analyzes of the microscopic, however relevant, components of each procedural class and observe
any legal divergences, whether temporal or sequential between the highlighted movements.
We emphasize that such methodology is commonly used in studies using complex networks.
The highlighted equations can serve as a guide for the creation of other exploration indices
of the network nodes that will quantify each type of procedural movement. Attributes such
as closeness, betweenness and pagerank have become essential to rank the efficiency of several
lawyers showing their importance in the community in which they work[12]. Thus, finding
information obstacles, distinguishing points of maximum power to disseminate information,
verifying possible waste and profits, are possible products arising from analyzes through complex
networks [13, 14].
4. Results
Thus, for each judicial class, we obtained the movements that characterize the most probable
steps (PP). In the Figure (2) we present a panel that represents the average time spent in each
type of movement for the different classes studied. For this step, we highlight the most common
points for constructing the paths that move 𝐹𝑖𝑗 to initiate and end a process. Thus, for the
Execução de Título Extrajudicial (Execution of Extrajudicial Title) class (Figure (2 (a))) the most
probable path indicates actions that lead to changes: Conclusão (Conclusion) → Mero expediente
(Mere expedient ) → Encaminhado edital / relação para publicação (Forwarded notice / relation
for publication ) → Despacho / Decisão disponibilizado no Diário de Justiça Eletrônico (Order /
Decision made available in the Electronic Justice Journal) → Trânsito em julgado (Unappealable
transit) → Definitivo (Definitive) → Expedição de Certidão de Arquivamento (Filing Certificate
Dispatch). It is noted that during the handling of the Mero expediente movement there was
the greatest expenditure of time. In the classes Common Civil Procedure, Special Civil Court
Procedure and Tax Enforcement (Figures (2 (b) - (d))) the types of preferred movements are
diverse. For the Procedimento Comum Cível (Common Civil Procedure) class, we observe the
similar sequence, representing accentuations in delays, in the Mere expedient and Trânsito
em julgado movements. Similarly, it is possible to observe in the Procedimento do Juizado
Especial Cível (Special Civil Court Procedure) and Execução Fiscal (Tax Foreclosure) classes
where the movements named as Expedição de Carta (Letter Dispatch) and Certidão emitida
(Issued Certificate) create long delays.
(a)
(b)
(c)
(d)
Figure 2: Average of delay time (Δ𝑇), in days, for various movements in classes (a) Execução de Título
Extrajudicial (Execution of Extrajudicial Title), (b) Procedimento Comum Cível (Common Civil Procedure),
(c) Procedimento do Juizado Especial Cível (Special Civil Court Procedure) and (d) Execução Fiscal (Tax
Enforcement). The start and end nodes of a process are indicated in the vertical part of each graph in its
ascending form. For example, in (a), a process begins with the movement of the Petition Protocol type
and ends with the Definitive movement.
The duration of a movement shown in the Figure (2) serves as a guide for the analysis of
each movement. In fact, movements have unique characteristics. Internal relations between
judges, lawyers, victims and defendants can influence the time of the movement depending
on the class to which it is contained. For example, the class Execução de Título extrajudicial
(Execution of Extrajudicial Title) deals with the procedure that aims to ensure the receipt of an
amount represented by an extrajudicial document (check, promissory note, contract signed by
two witnesses). The movements of this class contain activities that involve losses and damages
from both parts of the process, lead information of fiduciary alienation, relationships between
contracts, whether civil or bank, and even movements that contain contexts involved with the
environment. Some movements are internal to the judicial system. Others depend on external
public departments. In this case, depending on the actions involved and the history of the
process, it is possible to have a long delay in movement where causal stretches that depend on
police investigations or laboratory analysis can lead to an extreme time. Thus, the frequency
between movements tells us information about the pre-determined character imposed on the
local legal system, given by the flow chart of standard actions, how it carries cultural information
in the way it is resolved. Thus, knowing the activities that lead to movements to have such
different temporal patterns can be a way of verifying points that depend on greater attention
from the judiciary or just a way to reclassify procedural movements in a more convincing
way with the local reality. Figure (3) shows the time distribution density curves 𝑔(Δ𝑇 ) for the
most important procedural movements of each class presented in this work. Note that the
distribution pattern is different for each class. There is no defined shape for the curves and what
reinforces the idea that each movement can be studied in detail to evaluate the peaks found
in 𝑔(Δ𝑇 ). Note that in the classes Execução de Título Extrajudicial (Execution of Extrajudicial
Title), Procedimento Comum Cível (Common Civil Procedure), Procedimento do Juizado Especial
Cível (Special Civil Court Procedure) the moviment Trânsito em julgado (Unappealable transit)
is presented with similar temporal characteristics (see Figure(3 (a)-(c))). This movement is
used to identify the moment when a decision - sentence or agreement - becomes definitive,
and can no longer be appealed. There is a modal peak at Δ𝑇 = 20 days indicating a higher
temporal frequency for the time spent on this movement. However, the other fluctuations
may be the result of legal actions involving details of each class because, even though they are
common in both, Traffic accidents (commonly seen in Common Civil Procedure) and Property
Tax (observed in Execution de Extrajudicial Title) are, of course, distinct problems that involve
different actions but are computed, in some temporal phase, by this type of movement. Thus,
we reform that the diverse peaks in 𝑔(Δ𝑇 ) is caused by the divergence between the diverse
subjects assigned within the movements, and also within each class, which indicates that a
reclassification of the nomenclature of the states involved may be necessary. within a legal
process.
It is important to mention that the number of movements listed in the most probable path
for each process class depends on the choice of their start and end milestones. Statistically, the
most influential peers stood out from the rest. In addition, it is possible that there are paths with
a high degree of similarity due to common movements, even on routes of more or less stages.
Thus, for each possible pair of starting and closing a process, within a given class, we calculate
the variables 𝑁𝑚 and 𝑇 = ∑𝑘 Δ𝑇𝑘 which are, respectively, the number of moves found to obtain
the most probable path between network nodes and the total time spent on that path for each
(a) (b)
(c) (d)
Figure 3: Time distribution density curves 𝑔(Δ𝑇 ) for (a) Execução de Título Extrajudicial (Execution
of Extrajudicial Title), (b) Procedimento Comum Cível (Common Civil Procedure), (c) Procedimento do
Juizado Especial Cível (Special Civil Court Procedure) and (d) Execução Fiscal (Tax Enforcement) classes.
Each curve represents the most important movements linked to the movements shown in Figure (2). 𝑇 is
express in days.
step immersed in it that took Δ𝑇 time. For example, as shown in Figure (2(a)), chosen as starting
and closing points, respectively, the Conclusão and Experdição de Certidão de Arquivamento
movements, we have an amount of 𝑁𝑚 = 7 movements and a total of 𝑇 = 158 days.
The Figure (4) shows the scatter plots between such variables. Note, again, that the time
scale and the amount of movements required to obtain the most probable path depends on the
choice of the studied class. It is important to note that the relationship between the variables
suggests a linear correlation in the form 𝑇 = 𝜆𝑁𝑚 + 𝑏 [15] where 𝜆 indicates, on average, the
increase in time allocated to a process to a unitary variation to the number of movements of
that specific class. The lines highlighted in the Figure (4) indicate the linear adjustment between
the variables with Pearson’s coefficient [16] greater than 0.72 for any type of class cited.
The values of 𝜆 obtained by the linear adjustment for the classes Execution of Extrajudicial
Title, Common Civil Procedure, Procedure of the Special Civil Court and Tax Enforcement, are,
respectively, 21.3 days / movement, 11.5 days / movement, 11.9 days / movement and 26.4 days
/ movement. Its values reflect, graphically, the slope of the line in the graphs of the Figure (4)
taking into account the dispersion of points in the graph. Thus, for each class, it is possible to
estimate the effects of delays caused by the increase in the number of transactions and to rank
(a) (b)
(c) (d)
Figure 4: Average time spent 𝑇, in days, by a process obtained by the most probable path. Each point
indicates a combination of a possible initialization movement and another possible closing movement,
generating a sequence of 𝑁𝑚 total movements. The results refer to the classes (a) Execução de Título
Extrajudicial (Execution of Extrajudicial Title), (b) Procedimento Comum Cível (Common Civil Procedure),
(c) Procedimento do Juizado Especial Cível (Procedure of the Special Civil Court) and (d) Execução Fiscal
(Tax Enforcement), respectively. The highlighted lines refer to the linear adjustments in the form
𝑇 = 𝜆𝑁𝑚 + 𝑏.
those classes that have the highest speed in closing a legal process.
5. Conclusion
The developed article contributes to the development of research in the area of construction
and evaluation of networks in the legal context, at the same time that it has the potential to
allow the production of easily dated structures capable of guiding the implementation of public
policy and correction actions aimed at increased productivity and better provision of justice.
The methodology used for the creation and evaluation of the data generated by the procedural
movements is configured as innovative due to two important factors. The first is that it uses an
approach combining data analysis and network construction with the legal context, an approach
with little reference in the computer literature as in the legal or legal engineering literature. The
second is because it observes and values a source of information that is normally overlooked and
little valued in studies dedicated to understanding the characterization, dynamics and efficiency
of legal proceedings.
Finally, the set of data generated by the characterization of the procedural movement networks,
especially the inferences generated by the visualization of the time intervals between movements
and also the prediction of the duration of the processes, in view of the flow of movements
that they take, allows managers of the judiciary access to important data to both compare
performance levels between different jurisdictional units as well as to verify the occurrence of
efficiency ’bottlenecks’. A good example of this potential is the observation that in the case
159 - Execution of Extrajudicial Title ”(Figure (2 (d))), the simple dispatch of a letter (Expedição
de carta moviment) takes on average more than 50 days, the which indicates the need for the
adoption of corrective measures so that the provision of jurisdiction is faster and more efficient.
6. Acknowledgments
Acknowledgments
We would like to thank the Ceará Foundation for Support for Scientific and Technological
Development (Funcap) for the financial support in addition to the partnership with the Ceará
State Court of Justice (TJ) that provided us with the data for this project.
References
[1] M. Koniaris, I. Anagnostopoulos, Y. Vassiliou, Network analysis in the legal domain:
a complex model for European Union legal sources, Journal of Complex Networks 6
(2017) 243–268. URL: https://doi.org/10.1093/comnet/cnx029. doi:1 0 . 1 0 9 3 / c o m n e t / c n x 0 2 9 .
arXiv:https://academic.oup.com/comnet/article-pdf/6/2/243/24452305/cnx029.pdf.
[2] M. E. Newman, Mixing patterns in networks, Physical review E 67 (2003) 026126.
[3] A.-L. Barabási, E. Bonabeau, Scale-free networks, Scientific american 288 (2003) 60–69.
[4] X. F. Wang, G. Chen, Complex networks: small-world, scale-free and beyond, IEEE circuits
and systems magazine 3 (2003) 6–20.
[5] S. Boccaletti, V. Latora, Y. Moreno, M. Chavez, D.-U. Hwang, Complex networks: Structure
and dynamics, Physics Reports 424 (2006) 175–308. URL: https://www.sciencedirect.com/
science/article/pii/S037015730500462X. doi:h t t p s : / / d o i . o r g / 1 0 . 1 0 1 6 / j . p h y s r e p . 2 0 0 5 . 1 0 .
009.
[6] L. Amaral, J. Ottino, Complex networks. augmenting the framework for the study of
complex systems, European Physical Journal B 38 (2004) 147–.
[7] J. B. Ruhl, D. M. Katz, M. J. Bommarito, Harnessing legal
complexity, Science 355 (2017) 1377–1378. URL: https://science.
sciencemag.org/content/355/6332/1377. doi:1 0 . 1 1 2 6 / s c i e n c e . a a g 3 0 1 3 .
arXiv:https://science.sciencemag.org/content/355/6332/1377.full.pdf.
[8] Y. Lupu, E. Voeten, Precedent in international courts: A network analysis of case citations
by the european court of human rights, British journal of political science 42 (2012)
413–439.
[9] C. S. J., The network structure of supreme court jurisprudence. university of houston law
center, SSRN (2005).
[10] A. A. Pinto, S. R. de Souza, SoluÇÃo do problema de roteamento de veÍculos com janela de
tempo via iterated greedy search, Revista Interdisciplinar de Pesquisa em Engenharia 2
(2017) 182–195. URL: https://periodicos.unb.br/index.php/ripe/article/view/15042. doi:1 0 .
26512/ripe.v2i9.15042.
[11] E. W. Dijkstra, et al., A note on two problems in connexion with graphs, Numerische
mathematik 1 (1959) 269–271.
[12] L. Ribeiro, D. Figueiredo, P. Nascimento, Análise e ranqueamento da rede de advogados
induzida por processos judiciais trabalhistas, in: Anais do V Brazilian Workshop on Social
Network Analysis and Mining, SBC, Porto Alegre, RS, Brasil, 2016, pp. 13–24. URL: https:
//sol.sbc.org.br/index.php/brasnam/article/view/6441. doi:1 0 . 5 7 5 3 / b r a s n a m . 2 0 1 6 . 6 4 4 1 .
[13] C. Caminha, V. Furtado, V. Pinheiro, C. Ponte, Graph mining for the detection of over-
crowding and waste of resources in public transport, Journal of Internet Services and
Applications 9 (2018) 1–11.
[14] C. Caminha, Modeling user reports in crowdmaps as a complex network, 2012.
[15] W. H. Press, S. A. Teukolsky, W. T. Vetterling, B. P. Flannery, Numerical Recipes in C,
second ed., Cambridge University Press, Cambridge, USA, 1992.
[16] D. Freedman, R. Pisani, R. Purves, Statistics (international student edition), Pisani, R.
Purves, 4th edn. WW Norton & Company, New York (2007).