An Improved Ad Hoc Approach based on Active Help Method to

Detect Data Flow Anomalies in a Loop of Business Process Modeling

N. Chadli, M. I. Kabbaj

and Z. Bakkoury

Mohammed V University, Avenue des Nations Unies, Rabat, Morocco

Mohammadia Engineering School, Avenue Ibn Sina, Rabat, Morocco

Keywords: Data Flow modeling, Active Help, Loop Modeling, Verification, Validation.

Abstract: The data flow in business process modeling is created and distributed by the exchange of data moving from a

task to another in information systems. Among open issues in workflow modeling is the detection of errors

for data flow and control flow correctness. Recently researchers have focused on detecting errors by applying

an active help with a concept of Data-Record. However, this method does not support a loop modeling yet.

The goal of this paper is to apply an active help with a Data-Record concept in order to detect data flow

anomalies in loop modeling. We propose to improve the active help approach by some suitable Rules for loop

modeling. In this context, a decision node, using a data connection as an input data, replaces the connector

Xor-split. The input data of the decision node is returned to the last activity by a feedback when the error

message is found. The proposed approach is validated using a deterministic finite state process model which

uses a logic Boolean predicate (Yes or No) to specify the routing of an input data. Moreover, anomalies such

as Missing Data, Conflicting Data and redundant Data are used. The verification is triggered when an anomaly

is detected, where the system is locked until a correction is performed. The results show that Missing Data

anomalies are efficiently handled by the proposed approach.

1 INTRODUCTION

Currently, many business functions such as

purchasing, manufacturing, marketing, engineering,

and accounting have been automated by most

organizations, (A Basu, RW Blanning, 2000). To use

these functions, the data exchange in an information

system is necessary for each task to the next in the

business process management. In this sense, each task

requires input and output data. Speciﬁc collection of

tasks, resources and information elements make up a

workﬂow system as in (A Basu, RW Blanning, 2000).

Indeed, the business process activities are realized

throughout tasks in the information systems. Also,

business process activities can be achieved by

information systems without any human involvement

(TH Davenport - 1993 – books). In fact, in the

business process management, it is necessary to use a

workflow system in order to interplay between data

flow created by a data exchange of information

systems and the control flow in a workflow.

Consequently, data flow is important for business

process integration because data is always classified

when conducting inter-organizational business and

data errors could still happen even given syntactically

correct activity dependence. However, the focuses of

control perspective and data perspective to describing

the logical order of tasks and the information

exchange between tasks on verification is for most of

the techniques, i.e., on the discovery of design errors.

Certainly, the ﬂow-oriented nature of workﬂow

processes styles the Petri net formalism is a natural

contender for the modeling and analysis of workﬂows

(LIU Cong, Q ZENG, D Hua ,2014). In the time of

processing, there are many issues found when a

continuous passage the data flow of a process models

from an activity to the next. Many solutions have

been proposed by researchers to resolve the data flow

anomalies, as each activity needs operational

information to define the state of data, that is Read,

Write or Destroy. Therefore, this operation can

specify the state of data in the activity to another that

can cause missing data, conflicting data and

redundant data (SX Sun, JL Zhao, JF

Nunamaker,2006). An approach ad hoc uses active

help (MI Kabbaj, A Bétari, Z Bakkoury, A Rharbi,

2015) is proposed in a linear model and Xor-split with

two branches. This approach helps as to verify in each

284

Chadli, N., Kabbaj, M. and Bakkoury, Z.

An Improved Ad Hoc Approach based on Active Help Method to Detect Data Flow Anomalies in a Loop of Business Process Modeling.

DOI: 10.5220/0009775402840290

In Proceedings of the 1st International Conference of Computer Science and Renewable Energies (ICCSRE 2018), pages 284-290

ISBN: 978-989-758-431-2

time the task and corrected the errors in the same time

used a locked system and a concept DataRecord (MI

Kabbaj, A Bétari, Z Bakkoury, A Rharbi, 2015).

Moreover, they didn't use the loop when the system

ad hoc had a problem in a send message in a linear

model or model with Xor branches. In this way, the

same approach is used when adding a loop modeling

in the linear model with an Xor split in order to detect

the data flow modeling anomalies. Indeed, the Xor -

split is used to feedback an existing message errors at

a proceeding of modeling, this message errors

returned to the source activity where is created it up

to proceed of correction. Therefore, employ active

help (MI Kabbaj, A Bétari, Z Bakkoury, A Rharbi,

2015), and the rules for verification in the model, that

is triggered when some issue in the time of modeling

has occurred. However, the loop couldn’t assimilate

this approach to detect the anomalies not because the

active help is insufficient but because the rules of this

approach could only create and update. Subsequently,

it's proposed to enrich this approach with some

enhancements in rules and model in order for the

approach to be adapted by the loop. A decision node

is proposed like a connector that has a data connection

at the input data. In this case, it requires a Boolean

predicate (Yes=true, No=false) in a finite-state

automaton determinist, so we used the guarding (i.e.

blocking) tasks solely on the DataState (N Trcka, W

van der Aalst, N Sidorova,2008). In this context, we

implemented DataState to verify the last record state

of the dataset for each input and output in the activity.

In this manner, this data connection is a decision

variable that is a routing decision can be made based

on a set of data items inputted to the decision node.

Each of such data items involved in a routing decision

is called a decision variable (SX Sun, JL Zhao, JF

Nunamaker,2006). Also, this decision variable is

allowed to change the state of DataState that can be

initialized in each iteration of connection. Moreover,

there isn't the problem in the first iteration however

when the iteration is high requires an initialization of

the DataState.

The remaining of the paper is organized as

follows. Section 2 presents some approach and

concepts used in this paper. Section 3 shows that the

loop modeling cannot integrate assimilate the

approach with active help. In Section 4 presents the

new visualization of the approach. Section 5

concludes the paper and discusses the perspective.

2 RELATED WORK

Modeling in the business process has become very

important in recent years, with data-flow modelling

and verification being the two important challenges in

workflow system management. It had many works

stakeholders in this problem of anomalies of data-

flow and control flow in the workflow. Recently, data

flow formalization in process modeling has been

investigated by many researchers. In most

organisation, it is particularly important that the

responsible of key processes feel their interests are

represented during the latter phase. To achieve this,

the main stakeholders such as the heads of key

functions intersected by the process, the managers

with operational responsibility for the process,

suppliers of important change resources (e.g., the IT,

human resource, and financial functions), and process

customers and suppliers, both internal and external

should participate in the team during the design

phase. (TH Davenport - 1993 – books). In graph-

based approaches to business process modelling, data

dependencies are represented by data flow between

activities. Each process activity is given a set of input

and a set of output parameters. Upon its start, an

activity reads its input parameters, and upon its

termination, it writes data it generated to its output

parameters. These parameters can be used by follow

up activities in the business process (M Weske p.100

,2012). The importance of data-ﬂow veriﬁcation in

workﬂow processes was ﬁrst mentioned in (S Sadiq,

M Orlo, W Sadiq, C Foulger, 2004). The information

perspective defines what data are expended and

produced with reverence to each activity in a business

process. Thus, the operational perspective requires

what tools and applications are used to execute a

particular task (SX S, JL Z, JF N, 2006). Many

approaches have been proposed for for data-ﬂow

veriﬁcation, these approaches enable systematic and

automatic elimination of data-ﬂow errors as in (SX S,

JL Z, JF N, 2006). An approach of ad hoc that treated

the anomalies of data-flow for each activity by an

active help using a conception of dataRecord which

stored data with their state read, write and destroy

presented in (MI K, A B, Z Ba, A R ,2015). Indeed,

data flow perspective approach formally discovers

the correctness criteria for data-ﬂow modeling. Petri

Net based approach is proposed for modeling the

control flow of workflow. We extended this model by

including the input and output of data flow and added

a complexity of algorithm for detecting the anomalies

of data flow as in (LIU C, Q Z, D H ,2014). Our

approach extends and generalizes data flow

verification methods that have been recently

An Improved Ad Hoc Approach based on Active Help Method to Detect Data Flow Anomalies in a Loop of Business Process Modeling

285

proposed. It also makes use of the concept of

corresponding pairs lately introduced in control flow

verification. It has, therefore, the potential to be

developed into a unified algorithmic procedure for the

concurrent detection of control flow and data flow

errors. The aim of this paper is to present an algorithm

called GTforDF, for data flow verification through

the detection of lost data. The paper also will explain

through practical examples how GTforDF detects

data flow errors in workflows and define an important

new error category called redundant data in loops that

can lead to data loss in some situations (HS Meda, AK

Sen, A Bagchi, 2007). The approach focuses on the

discovery of data flow errors in workflows such as

Redundant Data, Lost Data, Missing Data,

Mismatched Data, Inconsistent Data, and Misdirected

Data. To achieve this, we an analysis which uses

“The RWD Boolean Table Technique” that is

expressed in steps, to split data-flow from control

flow and to create Boolean table for each data

elements, and also to compare RWD Boolean table

for current task and next task until it gets to the end

of workflow as in (AE Rgibi, SZ Yao, JJ Xu,2012) .

A three-layer workflow model for designing a

workflow was proposed in (FJ Wang, CL Hsu, HJ

Hsu ,2006). They characterized the behavior of an

artifact by its state transition diagram and identified

six inaccurate usages affecting workflow execution

and a set of algorithms a set of algorithms to detect

these inaccurate usages in workflow specification is

presented (FJ Wang, CL Hsu, HJ Hsu ,2006). An

approach in data flow issues proceedings for mapping

BPMN to Petri-Net to provide a systematic technique

of possible flows related to the data flow of business

process Data flow issues and BPMN mapping to Petri

Net: Road map as (AES Rgibi,2015). In a nutshell,

the objective of this paper is merging the Decision

Node with an input data and a logic Boolean by the

guard to find a new solution to solve the problems of

data flow anomalies in the business process with a

loop modeling in a linear model and Xor split.

3 VERIFICATION APPROACH

WITH A LOOP MODELING

In the case where an ad hoc method is applied in

a simple linear model with a loop, the system is

triggered when an error message of transmission is

produced. So, a feedback is structured to return error

to correct it. Otherwise, the modeler continues to

execute the next task as in figure 1. So, the feedback

requires the verification to detect data flow anomalies

in each system workflow instance. Consequently, the

output data became an input data in an information

exchange system of data flow. Indeed, at the moment

of the error message is returned, the approach is

triggered, and the verification is required to detect the

anomalies in each system workflow instance. As a

result, in the feedback and in each activity the output

data becomes an input data in an information

exchange system. Additionally, in a workflow, each

activity performs a comparison operation on a data

element. Thus, data operations are spontaneous, when

an activity A is reading data, the item d is an input

data. The same, when an activity A carry out a writing

operation on d, d is the output data from A as in (SX

Sun, JL Zhao, JF Nunamaker,2006). Furthermore, it

is proposed to verify some anomalies such as Missing

data, conflicting data, redundant data for loop

modeling applying the approach rules (MI Kabbaj, A

Bétari, Z Bakkoury, A Rharbi, 2015).

Figure 1: Description of linear model and loop modeling.

3.1 Description of Symbols Used in the

Model Looping

The tables 1, 2 and 3 define and describe the

symbols and operations used in the model in figure 1.

Table 1: Description of Data item.

Data item Description

m Message1(email)

m’ Message2(email)

d destination

e error

a Accuse

ICCSRE 2018 - International Conference of Computer Science and Renewable Energies

286

Table 2: Description of activities.

Activities Description

A1 Write email

A2 Corrected email

A3 Send email

N1 Decision Node

Table 3: Description of Operation.

Operation Description

R Read

W Write

D Destroy

I Input

o Output

3.2 Verification the Model with a Loop

We apply the Rule 1 of approach (MI Kabbaj, A

Bétari, Z Bakkoury, A Rharbi, 2015) in the example

below in figure 2. Rule 1:

“For an activity, a given data d with the state

(x,y,z), if d is inserted for the first time in the

DataRecord and x ≠ 0 we have an error ==>

uninitialized data (missing data).”

The conclusion is that data item d is detected as

missing data in activity A2 in figure 2 & table 4

below.

Figure 2: Data Flow modeling in linear model.

Table 4: Simple state without iteration.

Data State

(0, W

,0)

m’

(0, W

,0)

The modeler chooses two solutions to correct the

anomalies of missing data, either not to read the data

item d in activity A2 (to destroy) or write data d in the

activity A2. Then, data d is not read in activity A2. In

this instance, before the first iteration, an error e has

occurred at processing in the activity A2.

Consequently, after drawing the model and using a

connector Xor-split to have the conditions for the

error e to occur. This model is considered as a finite

state determinist model with Xor-split as a node.

Indeed, if the error e is written in activity A2, the

feedback loop returns the message for correction,

otherwise, the process continues modeling. In fact, it

is required to verify the anomalies at modeling to

detect missing data in the two cases applying this

approach (MI Kabbaj, A Bétari, Z Bakkoury, A

Rharbi, 2015). Applying the Rule 1:

if e≠ ∅ it is noteworthy that the DataState

(DataRecord) contains the latest data set, and

activities as in table 5. At the verification, the loop

cannot analyse the anomalies of data flow existed in

processing, so no missing data is detected.

Table 5: The first iteration.

Data State

, W

,0)

m’

(0, W

,0)

(0, W

,0)

Otherwise, if e=∅, the system continues with the

modeling, and applies the Rules 4 and the process

verification continues as in table 6, and the

redundant data a is detected.

Table 6: The new dictionary with iteration.

Data State

, W

,0)

m’

, W

,0)

(0, W

, 0)

(0, W

,0)

We conclude, that for loop modeling with a connector

Xor-split, even if they are based on data this will not

change anything in this approach (MI Kabbaj, A

Bétari, Z Bakkoury, A Rharbi, 2015), as the Rules are

not practicable. Therefore, it’s proposed to enrich this

approach (MI Kabbaj, A Bétari, Z Bakkoury, A

Rharbi, 2015) to take into consideration the loop

when the error in the message has occurred.

4 DESCRIPTION OF THE NEW

APPROACH

4.1 Description and Definition

Dataflow is often defined using a model or diagram

in which the entire process of data movement is

An Improved Ad Hoc Approach based on Active Help Method to Detect Data Flow Anomalies in a Loop of Business Process Modeling

287

mapped as it passes from one component to the next

within a program or a system, taking into

consideration how it changes form during the process.

Thus, data item of data flow has two roles; one is data

link which is connects an activity to another by an

input data and an output data. The two roles are to

transmit the information from one task to another. In

our situation data has the two roles and this data is

extended by read and write/update and destroy.

Consequently, when data transmit information there

are many errors to be tackled such asMissing data,

conflicting data and redundant data.

4.1.2 Decision Nodes

The decisions node is a conditional construct which

can also be modeled with a conditional node using a

logic Boolean predicates with a function guard that

can allow us to model decision points in which the

choice is made based on some data elements. When

the model uses decision nodes, usually their edges

have guards that are Boolean logic value specification

evaluated at runtime to discover if control and data

can be evaluated along the edge. Additionally, for

each individual control and data token evaluated by

the guards at the decision node to get precisely the

edge that the token will be extended across. we can

say that a decision nodes are Task nodes that

represent atomic manual automated activities or

subprocess that must be completed to fulfil the below

business process objectives (SX S, JL Zhao, JF

N,2006).

4.2 The New Approach

The approach based on method ad hoc applying an

active help attended with the concept DataRecord for

verification would be improved in order to be

compatible with the loop. Consequently, it is

suggested to enrich this method in order to be able to

apply Rules on a loop. Additionally, an Xor-split is

used by the model as a connector to feedback the

exchange data flow when the error message has

occurred. Indeed, it is proposed to use the Xor-split as

a decision node with a data connector at an input data

I(d). Moreover, this data connection is a routing

decision which can be made based on a set of data

items inputted to the decision node for a function to

return the data when the feedback with a read

operation in the activity A2 in which the error

message occurred. But this proposition needs the

concept DataRecord that it changed by a DataState

which is to be initialized for each iteration and carry

the latest activity and state of a dataset. The

verification of detection data flow anomaly is the

same, we keep an active help in an ad hoc approach.

Indeed, each activity has an input data I(d) formalized

as (RA,0,0), d is read in the activity A, and an output.

O(d’) in activity A' as (0,wa',0) d' is writen in activity

A', e.g. if I(d) is accessed by activity A that produces

an output data O(d’) by activity A'. Otherwise, an

input I(d) has to be processed for the next task and

data will not updated. Thus, an input data I(d) is read

in the decision node and data d is chosen to be read in

the activity A2 in figure 3. In order to solve the above

problem in the precedent section, the Decision Node

N1 that is considered as an activity required by adding

an input data I(d) as a decision variable. Additionally,

a Boolean logic predicate (Yes, No) with a guard to

specify the routing of a Decision Node N1 that (Yes

=true & No=false) is added.

4.2.1 The Solution with a Decision Node

We began drawing the model for each task at the

moment of modeling applying the proposed

improvement of the approach. Therefore, if the error

e exists (e≠∅) the guard is "No", in this instance, the

system must be on feedback loop and read data d

(destination) in the activity A2 which has been I(d) in

the activity N1 that it couldn't be writen or updated in

A2. Otherwise, (e=∅) the guard is Yes, and no change

in the next activity A3 the model continues

processing. Consequently, the loops and Xor can only

read the data; it can't delete nor create nor update data.

In this case, the given data flow anomaly is only

missing data. In this case, the missing data rule would

be ameliorated, and the concept DataState manages

the state of each data in the activities which have been

reinitialised in each iteration, as in figure 3.

Figure 3: The conditional Node decision.

ICCSRE 2018 - International Conference of Computer Science and Renewable Energies

288

4.2.2 Rule Missing Data

For each iteration, DataState is initialized to record

the new stored data set and activity and the

decision node N1 has a guard No, the data d

is in read in activity A2, involving the

missing data.

if  → & , then d is read in the

next activity; therefore, involving missing data.

4.2.3 Rule of Message Error

When an error e is created in a task, an output O(e)

occurs, then the process cannot continue their task

until the next activity due to the error e.

Consequently, the loop starts, and if the guard is false

(No), the system is loaded to correct the error at the

next activity A2 (corrected email) in time of looping

by destroying it.

4.2.4 Interpretation

We suppose that the error e is corrected when the

system returns to activity A2 by applying Rule 2, and

the error e is destroyed. At the same time when the

preceding task is connected to node N1 by an input

data d, the system has a choice to return the message

by No to correct the error, that involve data d in

reading in activity A2. Applying Rule 1, the system

has detected an anomaly d in A2 missing data, after

verification and locking the system, the modeler

chooses to correct the anomaly with write data d in

activity A2 as in DataState in table 7.

Table 7: The new DataState with loop and Decision Node.

Data State Iteration

m (R

, W

,0) 1

m’ (0, W

,0) 1

e (0, W

, D

) 1

d (0, W

,0) 1

The decision Node has a connector input data I(d)

which is not depended on the number of iterations.

Consequently, our approach with a (rules 1 & rules 2)

are valid for n iteration:

when   1,   0 n iteration as in table 8.

Table 8: The DataState for n iteration.

Data State Iteration

m (R

, W

,0) n

m’ (0, W

,0) n

e (0, W

, D

) n

d (0, W

,0) n

4.3 Future Work

In the future, we will try to find a solution for the

other anomalies i.e conflicting data, redundant data

and to verify this method with a model checking to

verify of each instant the model by an active help at

the moment the anomalies are detected, a

computation tree logic CTL* in a temporal logic used

subset LTL is to be investigated, and to find an

example on which we can validate the model . (N Tr,

W van der Aal, N Si, 2008; EM C, O Gr, D P - 1999;

EA E JY Halper,1986).

5 CONCLUSIONS

The main focus of this paper is to use ad hoc approach

and active help and a concept of DataRecord to verify

data flow anomaly issues in loop modeling. In this

case, we ameliorate existing rules by adding a

decision node to make the grade by Yes or No. Also,

we reinitialize DataState in each iteration. As such we

have detected some anomalies of missing data, but we

could not detect the others anomalies i.e. conflicting

data and redundant data, because in the loop we can

only read but not create, or modify. Consequently, we

must ameliorate this approach in future work to detect

more anomalies of data flow modeling in the business

process.

REFERENCES

A Basu, RW Blanning - Information Systems Research,

2000 - pubsonline.informs.org A Formal Approach to

Workﬂow Analysis.

TH Davenport - 1993 - books.google.com Process

innovation: reengineering work through information

technology.

M Weske - Business Process Management, 2012 – Springer

Business process management architectures.

S Sadiq, M Orlowska, W Sadiq, C Foulger - Proceedings of

the 15th …, 2004 - dl.acm.org, Data flow and validation

in workflow modelling.

An Improved Ad Hoc Approach based on Active Help Method to Detect Data Flow Anomalies in a Loop of Business Process Modeling

289

SX Sun, JL Zhao, JF Nunamaker… - Information Systems

…, 2006 – pubso, nline.informs.org, Formulating the

data-flow perspective for business process

management.

LIU Cong, Q ZENG, D Hua - International Journal of

Science and …, 2014 - ijsea.com, Formulating the data-

flow modeling and verification for workflow: A petri-

net based approach.

MI Kabbaj, A Bétari, Z Bakkoury, A Rharbi - IJCSA, 2015

- researchgate.net, Towards an active help on detecting

data flow errors in business process models.

HS Meda, AK Sen, A Bagchi - 2007 - papers.ssrn.com,

Detecting data flow errors in workflows: A systematic

graph traversal approach.

AE Rgibi, SZ Yao, JJ Xu - Applied Mechanics and

Materials, 2012 - Trans Tech Publ Dataflow Errors

Detection in Business Process Model.

FJ Wang, CL Hsu, HJ Hsu - Computer Software and …,

2006 - ieeexplore.ieee.org, Analyzing inaccurate

artifact usages in a workflow schema.

AES Rgibi - … and Industrial Engineering: Proceedings of

the …, 2015 - content.taylorfrancis.com, Data flow

issues and BPMN mapping to Petri Net: Road map.

N Trcka, W van der Aalst, N Sidorova - Computer science

report, 2008 - research.tue.nl, Analyzing control-flow

and data-flow in workflow processes in a unified way.

EM Clarke, O Grumberg, D Peled - 1999 -

books.google.com, Model checking.

EA Emerson, JY Halpern - Journal of the ACM (JACM),

1986 - dl.acm.org, on branching versus linear time

temporal logic.

ICCSRE 2018 - International Conference of Computer Science and Renewable Energies

290