=Paper=
{{Paper
|id=Vol-3295/paper16
|storemode=property
|title=Development of a Reflective Intelligent Project Planning System
|pdfUrl=https://ceur-ws.org/Vol-3295/paper16.pdf
|volume=Vol-3295
|authors=Iurii Teslia,Ganna Klevanna
|dblpUrl=https://dblp.org/rec/conf/itpm/TesliaK22
}}
==Development of a Reflective Intelligent Project Planning System==
https://ceur-ws.org/Vol-3295/paper16.pdf
Development of a Reflective Intelligent Project Planning System
Iurii Tesliaa, Ganna Klevannaa
a
Cherkasy State Technological University, Shevchenko blv., 460, Cherkasy, 18006, Ukraine
Abstract
Planning is one of the main functions of project teams and one of the most important
prerequisites for project success. Despite the development of information systems and
technologies, planning methods have remained virtually unchanged, including the critical
path method and the PERT method. They are based on the implementation of the opinion of
experts on the possibility of performing work in a timely manner, the need for resources, the
technological links between the works. It is shown that for the creation and implementation
of information systems and project management technologies it is necessary to create
methods and tools that will significantly transfer the planning functions to specialized
intelligent systems. The problems that should be solved by such systems are formulated.
Scientific and methodological approaches to the creation of information systems and
planning technologies are analyzed. It is shown that the reflex approach allows to combine
and model the influence on decisions in the process of project planning of different physical
factors - expert opinions, documentation, information standard of the company, the situation
in the project environment. We have utilized a modified reflex method to build an intelligent
project planning system. A modified reflex method of developing reactions to the influence
of various factors that determine the parameters of the project plan is proposed. The structure
of the reflex system of project planning is proposed, which is able to adequately respond and
implement decisions on the parameters of project work. The system has passed experimental
testing and has shown its effectiveness in the project planning process.
Keywords 1
Project planning, project management software tools, project management information
technologies, reflex approach
1. Introduction
Considering that planning is one of the main functions of project teams and one of the most
important prerequisites for project success, this process attracts and requires a lot of attention. Today,
the work of project management groups on their planning is impossible without the use of such
software tools as Oracle Primavera P6, MS Project, Jira, Trello etc. [1-3]. Despite this, planning
methods have remained largely unchanged, including CPM and PERT methods. There are two main
tasks in the planning process using software tools. Preparation of the Project Schedule Network
Diagram (PSND), taking into account the duration of the works, the amount of resources for the
works, the connections between the works. And its calculation with the receipt of the project plan.
The second problem is solved by instrumental software [1-3]. After all, these and other software tools
calculate network graphs, build Gantt charts, distribute resources depending on the time of project
implementation, etc. At the same time, they implement technical functions - calculate the network
schedule, shift work according to the performance results, calculate the loading of labor resources,
etc.
Proceedings of the 3nd International Workshop IT Project Management (ITPM 2022), August 26, 2022, Kyiv, Ukraine
EMAIL: dr.teslia@gmail.com (Iurii Teslia); h.o.klievanna.asp21@chdtu.edu.ua (Ganna Klevanna)
ORCID: 0000-0002-5185-6947 (Iurii Teslia); 0000-0002-7450-379X (Ganna Klevanna)
©️ 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 (CEUR-WS.org)
The first task is still solved by managers. Creative tasks related to the preparation of the PSND, in
particular, related to the development of the topology of the network model, the distribution of
resources, and the determination of the duration of work remain with the project management group.
Which places increased demands on the professionalism of managers in the tasks of project planning.
As experience shows, a significant number of projects are not completed in accordance with the
work plan. And this is not only due to poor execution of work. But also poor project planning,
primarily with poor PSND training. Therefore, the creation of intelligent project planning systems
corresponds to these trends and is very modern and urgently needed.
Today, there are no sufficiently effective expert systems that would interact with managers in
natural language and solve the tasks of project planning, allocation of resources between works,
execution control, and evaluation of the prospects of project implementation in the task of deadlines.
Although there is already work in this area to solve local intellectual problems, such as decision
support and intelligent data analysis. Thus, the papers [4-5] present the prospects of integrating
artificial intelligence technologies into project management. But the works lack the development of
intelligent systems for the implementation of the proposed models and methods.
This paper [6] presents an intelligent decision support system in the field of organization of non-
profit projects. This system helps determine whether a planned project will be successful by assessing
the likelihood of future project success, and provides advice and guidance on how to improve project
organization when the probability of project success is low. In work [7]) the information and
analytical system of management of the project team in software development is described. This
allows you to form effective teams using the Kanban methodology within software development
projects.
There are works devoted to the use of expert systems in the management of enterprises and
business environments [8]. But their scope of application does not overlap with the tasks of project
planning, so it is impossible to apply them in this area.
The analysis of literary sources showed that today there are no functionally complete intellectual
systems aimed at the formation of project plans. Who would be able to make effective decisions in the
process of forming and calculating project plans, allocating resources between works, and monitoring
execution. Therefore, there is an unrealized part of research in the field of creating project
management information technologies. This work will be devoted to the issue of creating an
intelligent project planning system that would adequately respond to various influences in the project
environment and would be able to draw up a project plan that would correspond to the realities of the
project.
2. Materials and methods
Our goal of the work is the development of an intelligent project planning system capable of
forming the PSND of the project with minimal human involvement. To achieve this goal, it is
necessary to the following tasks: to propose a method that would allow to form the PSND of the
project efficiently and simply, to develop an intelligent project planning system, and to conduct
experimental studies.
The object of research in the work is project planning processes.
The subject of the research is an intelligent system that will take on creative tasks to prepare a
project plan - the formation of PSND in terms of links between works, their duration and the
necessary resources.
The result of the study should be an intelligent planning tool that allows you to form a CG in the
process of verbal interaction with the project team.
2.1. Creative tasks that appear in the process of project planning
In the process of planning projects, the following functions are employed:
1. Molding for the development of the project plan.
2. Selection and analysis of information (documentation, confirmation of the implementation of
future projects, minds of the implementers of the project) for the development of the project plan.
3. Formation of PSND.
4. PSND calculation and obtaining a preliminary project plan.
5. Evaluation of the project plan.
6. Agreement and approval of the project plan.
7. Bringing the project plan to completion.
From rescheduling tasks to creative tasks, you can add tasks 1, 2, 3, 4, 5. Let's look at these tasks
in more detail:
Task 1. This is decided by the top management of the company and is related to the strategy of its
development and business activities and cannot be effectively solved in an intelligent system. Because
in the process of solving this, qualitative assessments, confidential information, the attitude of top
management and counterparties to the scope of project implementation are often used. But in the
future, this task can also be transferred to the digital environment.
Task 2. Applied by the information specialists of the project management group, as well as by the
practitioners of the company's functional developments. With the use of two models of project
planning (bottom-uphill, top-down), this task can be achieved with a different level of creativity of
specialists and practitioners. When planning from the bottom up, specialists and practitioners appoint,
as it is necessary to work out (in the area of their competence) and, apparently, pick up documentation
for “their expertise”. When planning from top to bottom, specialists and practitioners already take
away the preparation of the project and pick up information on the tasks, work is carried out with the
method of assigning their parameters (resources commitment, time spent, technological
communications with other robots). The task can be solved without human participation only if the
digital environment contains all the information about the project: the information standard of the
company (which reflects the previous experience of implementing projects), documentation, and an
expert assessment of the project and its implementation environment. The development of digital
technologies, the creation of methods and means of digital project management will sooner or later
lead to the solution of this problem without (or with minimal) human participation.
Task 3. Operate in the automatic creation of PSND based on the knowledge of experts (intellectual
systems), information standard, documentation (which determines technological links). As a matter of
fact, technological links can be removed not only from the documentation, but also from the
advancement of the implementation of the projects and the knowledge of the experts, and the task
today can be implemented in the intellectual system.
Moreover, the knowledge of experts is presented through the formalization of qualitative
knowledge based on an interview, the essence of which is to obtain answers to the questions posed.
For example: What will be the duration of X's work? What resources and in what amount are needed
to perform X's work? After which job(s) does job X start and with what delay? If work X starts later
than planned, what amounts of resources are needed to reduce its duration by P percent? This
information, together with the information generated when solving problem 2, is used to determine
the sequence of execution and parameters of project work.
Task 4. It is solved by instrumental project management software.
Task 5. Typically, formal methods are used to evaluate the project plan, in particular, calculations
within the project triangle (time, money, quality). These calculations show whether or not the
resulting plan will allow the project to be completed within the prescribed time and budget. And
whether this time, funds, as well as the planned loading of labor resources is enough to ensure the
required level of product quality of the project. In modern project planning systems, this approach is
quite productive and of high quality. Its disadvantage is insufficient consideration of the risks of non-
performance of the project within the planned funds and time. The use of intelligent systems to solve
this problem is possible through the introduction of qualitative assessments of the project plan
(reliable, low risk, rational in time and cost, rhythmic, etc.). This problem can be solved in the future
by the technologies from which problem 3 will be solved.
Based on the description of these tasks, we can conclude that the creation of an intelligent project
planning system with minimal human participation should begin with solving problem 3. Creating a
model of work based on expert knowledge, project conditions, documentation and information
standards of the company
In addition, it is necessary to assess the input information and the compliance of the task with the
capabilities of the project team and its resources. This will be done by displaying the results of solving
tasks 1 and 2 in the intelligent project plan evaluation system.
But for its creation, it is necessary to develop a method that will design the topology of the
network model and determine the parameters of the projects. Thus the system which will implement
this method should:
Receive expert information not only through the knowledge base, but also receive it in dialogue
with experts in the process of forming a work plan. To do this, it must be able to communicate with a
person on the subject of creating a PSND in a human language.
Take into account information from the information standards. Such information should influence
the decision of the intelligent system not only regarding the topology of the network model and
parameters of the work, but also about the trustworthy and reliable information from specific experts
(how wrong they were in previous projects to assess the parameters of the work).
Take into account the conditions of the project, described in the documentation. This information
specifies the data from the information standard by specific (and often new) resources, requirements,
limitations. And it is presented both in the human language form, and in drawings, tables, lists,
calculations. Therefore, the intelligent system must take into account the impact on the project plan of
heterogeneous information describing the conditions of the project.
It is very difficult to create an intelligent system that would be able to solve the formulated
problems. After all, these tasks use heterogeneous information. This makes it quite difficult to create a
single core for their solution, either on the basis of a neural network or with the use of knowledge
bases. To create intelligent systems that are able to combine in one process disparate information,
simple and cheap to build, easy to perceive and learn to respond properly to human language has long
used the reflex approach. This allows you to create simple and effective systems responsive to the
flow of heterogeneous data by developing reflexes to the informative component of such flows [9].
The main feature of reflex systems is the possibility of simultaneous processing of different types of
input data, from separate sources as a single input stream with the production of reflexes to arbitrary
combinations of elements of this data. It is most suitable for processing incoming data flow from
managers, formalized and human language information from project documents, statistical
information from information standard and regulatory information of project planning management.
Consider the problem of developing a response to the flow of arbitrary input data in solving
intellectual problems in project planning systems.
2.2. The structure of the reflex intelligent project planning system
The main task of the reflex intelligent system of project planning (RISPP) is to create a topology
of the network model PSND, to determine the parameters of work and resources. This task is
performed on the basis of information about the content of the project and the physical scope of work.
To solve this problem, the project planning system must: receive information on the content of the
project and the physical scope of work; use the information standard of the project-oriented company
to take into account previous experience in planning and implementation of project plans to set /
adjust the duration of work and resources; prepare information for software that performs the function
of calculating PSND (e.g., MS Project). To do this, the intelligent system must, in accordance with the
information coming into the system, form an array of data for the calculation of PSND. The
application of the reflex approach allows for different sets of input data to determine the
corresponding elements of the data set for the calculation of PSND by generating reflexes on
individual elements of this data.
Based on this, the reflex intelligent project planning system can be represented by the structure
shown in Figure 1. The key modules are two: the reflex generation module and the reaction
generation module. Consider methods of producing reflexes and developing reactions that are based
on a reflex approach to building intelligent systems.
Figure 1: Macrostructure of the reflex intelligent project planning system
2.3. Development of reflexes for the flow of input data in RISPP
Under the RISPP reflex we will understand the stereotypical reaction of the intelligent system to
the input data stream, which reflects the statistical regularity of the correct response of the system to
individual elements of this input data stream generated by the instructor in the process of machine
learning. Thus, the development of reflexes is the accumulation of statistical information on the
elements of the input data and the reactions that were identified by the instructor when these elements
appeared.
The illustration of the reflex can be seen in the following example. The correct reaction of the
project planning system to the appearance of the element in the input data stream: The milestone
"Start of the project" is the introduction of the PSND work "Start of the project" with a duration of 0
days. If several such elements appear, the system produces a reflex: If there is a word "milestone" in
the input data stream, the duration of such work will be 0 days. Or when texts appear that contain the
phrase "start work A after work B", a reflex is produced on the combination of the words "start…
after…", which consists in the reaction "connection between works - finish-start".
At the heart of reflexes is statistical information on the causal relationship between the input data
stream and possible reactions. At the heart of this statistical relationship is the conditional probability:
, (1)
where Rj is the reaction of RISPP; E - input data stream.
If there are several possible reactions, the RISPP should choose the one that has the highest
conditional probability
(2)
.
In general, the input data flow is practically not repeated in project planning systems. Accordingly,
it is impossible to accumulate statistics to select the most likely reaction (formula 2). Therefore, in the
project planning system, it is proposed to divide the input data stream into elements that can be
repeated in different input streams, and to produce reflexes on these elements. With the choice of the
reaction to which the reflexes are strongest.
The input data stream in accordance with the presented structure (see Figure 1) includes:
1. Text in human language from members of the project management teams, which determines the
reactions of the RISPP. In essence, the elements of such an input data stream can be sentences,
words, letters. But it is impossible to collect the statistics necessary for learning from repeated
sentences. It would be best to use words. On the one hand, the conditional probabilities of certain
reactions to words can be quite significant. But on the other hand, modern speech recognition
systems. But on the other hand, modern speech recognition systems quite often make mistakes in
the interpretation of words, and even managers use different terms to define the same reactions.
Therefore, it was decided in RISPP to use letter combinations of different (from 2 to 10 characters)
length. This increases noise immunity (you can make mistakes in words, then the reaction will be
formed on the part of the sentences that are written without errors).
2. Project documentation (for construction - DBNs, design and estimate documentation, contracts,
etc.). In RIX information is presented in two ways:
2.1. In human language form (the method of processing such an input data stream is discussed in
claim 1).
2.2. In the form of a frame. Where the name of the slot specifies the project parameter, and the
value of the slot is recorded as a data element from the documentation. In this case, the RISP
produces reflexes for the frame name and the slot name. Example:
{Frame name: trench digging work;
slot name 1: volume, slot value 1: 100 m3;
slot name 2: duration, slot value 2: 3 days;
slot name 3: resource, slot value 3: JCB JS175W full-turn excavator;
…
}.
The data elements for RISPP are the frame name and slot names. Reaction - setting the value of
the corresponding frame name (work) and the name of the parameter slot. In the example of the
reflex on behalf of the frame "trench digging work" - to set the parameters of work. Reaction to
frame names - definition of work for establishment of values of the parameters set in slots.
3. Data of the information standard. The information standard is supplemented with data on
implemented projects and is statistical information on the deviation of the fact of implementation
from the plan in previous projects. For use in RISPP data of works and resources are presented in
tabular form (Table 1 - table 2).
Table 1
Representation of the input flow of data for works on the information standards
Project name: Logistics center construction project
ID Work Planned duration Actual Connections
Expert 1 Expert 2 … Expert N duration (ID, Type,
Delay)
7001 Obtaining a 30 45 60 78 7000, FS,
construction permit 0
Table 2
Representation of the input flow of data for works on the information standards
Project name
Work Resource Planned amount Actual amount
ID Expert 1 Expert 2 … Expert N
7001 Representative 10000 5000 12000 5000
expenses
2.4. Development of reactions in RISPP
The response of RISPP is to form a data element or perform an action to calculate the project plan.
The reaction must correspond to the content of the input data stream. The reactions of the reflex
intelligent system to the input data stream can be:
1. WBS project structure
2. Project resources.
3. Project work, the relationship between them and the parameters (time, cost, volume).
4. Action to calculate the project plan.
The choice of reaction is based on the assessment of the total conditional probability of possible
alternative reactions to the input data stream. The reaction that has the highest probability is selected
based on the reflexes produced in the RISPP.
We estimate the total conditional probability of possible alternative reactions to the input data
stream as a reflex response to stimuli in living beings. If reflexes are made on separate stimuli, and
many influences (many stimuli) are exerted on a living being, then the strongest reflexes "work". For
RISPP, this rule can be paraphrased as follows. If reflexes are made on data elements, and the input
data stream consists of many elements, the reaction which corresponds to the strongest reflexes is
chosen, it “triggers” the strongest reflex. To implement this rule, it is necessary to be able to assess
the "strength of the reflex". In the reflex approach, such a force is determined by the difference in the
conditional probabilities (in the presence of influence, and without influence) of the probability of
reactions. Namely [9]:
1.
(3)
,
where is the increase in the effect on the reaction associated with the appearance in the input
data stream of the element ; – conditional probability of reaction in the presence of the
element e_i in the input data stream; – conditional probability of reaction in the absence
of the element in the input data stream.
In the general case, if there is a significant number of elements of the input data stream
(and for the planning system it is), you can take
,
where is the unconditional probability of the reaction .
2. The total increase in the influence of all elements of the input data stream on the reaction :
(4)
where is the total increase in the effects of all elements of the input data stream on the reaction
; E - input data stream.
3. The overall effect on the reaction
where is the value of the total influence of all elements of the input data stream on the reaction
.
4. Estimation of the probability of choosing the reaction , which corresponds to the
magnitude of the influence of the input data stream:
, (5)
where is an estimate of the probability of choosing the reaction , if the system receives
the input data stream E.
The choice of reaction is performed in accordance with formula (2).
As can be seen from this representation of the modified (for the production of reflexes on the input
stream of inhomogeneous data) reflex method, it is quite simple. Its effectiveness has been proven by
both practical developments and experimental studies [9]. Thus, it is this method that will form the
basis of project planning with a reflexive intelligent system.
2.5. Software and information environment of the reflex intelligent
project planning system
In subsections 2.3 and 2.4 the input and output information of RISPP is considered. Processing of
input data streams with the production of adequate information contained in these reaction data is
performed according to the following generalized algorithm:
1. Determination of the content of the input data stream (what meaningful information it
carries).
2. Determination of a meaningful scenario of actions (algorithm implementation of the content
embedded in the input data stream).
3. Distribution of fragments of the input data stream by elements of the meaningful action
scenario.
In the process of modeling project planning actions, 26 meaningful scenarios were developed, which
could be implemented in 113 action scenarios (an example of meaningful action scenarios is given in
Table 3).
Table 3
An example of meaningful action scenarios
Text Content A meaningful action scenario
Add the task of obtaining data Add a task to a group +task; task; +in;task
for writing a technical task for
performing topo-geodesic
studies to the Topo-geodesic
studies node
Start the Concept Expertise Communication "finish-start" +start; task; +after; task
task after Developing a New
Concept
Add resources to the node Add resources to the task +resource; task
Selection of a contractor by
shooting range
Based on the presented generalized algorithm for developing reactions to the flow of input data,
and taking into account the need to develop reflexes for input flows of information with different
purposes and their fragments, we offer the following structure of the information base and software
tools.
The information database will contain two groups of tables: tables of intelligent modules for
generating reflexes and developing reactions and tables of the project plan. Three tables are used to
display reflexes: U_S, U_A, U_R. The UML diagram of the database is shown in Fig. 2. The purpose
of the tables and the description of the fields are given in table 4.
Figure 2: UML diagram of the local RISPP database
Table 4
Representation of the input flow of data for works on the information standards
Table Field Type Purpose
U_S - input elements NzS Numeric Record ID
Fon Nvarchar An input data stream element
Kol Integer How many times occurred in the
input data stream
R Double The amount of heterogeneity of
alternatives
K Integer The number of alternative
reactions when an element appears
NzR Numeric Record ID
U_R - reactions Kol Integer How many times occurred in the
input data stream
Txt Nvarchar Reaction
D Double Average effect on response
U_А is the connection NzA Numeric Record ID
between U_S and U_R Nzp_1 Bigint The element ID of the input data
stream
Nzp_2 Bigint Reaction ID
Kol Integer How many times was this reaction
when this element appeared
D Double The magnitude of the effect of the
element on the reaction
(formula 3)
To increase speed and ensure the possibility of parallel processing of information with different
purposes in the system, a separate physical database was created for each class of reactions. These
were located independently of the others on a physical medium or in the Cloud. Moreover, the
structure of each database corresponded to the diagram shown in Fig. 2. These databases were not
connected to each other, as they contained different information for the production of different classes
of reactions.
A total of 12 databases were created by type of reactions (which contain the tables shown in Fig.
2) for the formation of reflexes with the development of reactions to control the planning and
formation of PSND:
Database 1. Determining the content of the action in the scheduling system, which is set by the
input data stream (BA_00). Table U_R of this database contains meaningful action scenarios. Content
options are shown in Table 4.
Database 2. Defining a meaningful action scenario (BA_01). In Table U_R of this database there
are options for the content of actions. An example of meaningful action scenarios is given in table 4.
Database 3. Relationship of fragments of the input data stream with the elements of the meaningful
action scenario (BA_02). Table U_R of this database shows the relationship between fragments of the
input text and the content of actions. It is used to select one of the following databases for processing
fragments of the input text (for example, job or resource name, job duration, etc.).
Database 4. Project works (BA_03). The reactions in this database are project works.
Database 5. Project resources (BA_04). Project resources are added (in the learning process) to
Table U_R as reactions.
Database 6. Projects (BA_06). In Table U_R - company projects.
Database 7. Duration of works (BA_07). Reactions - how many days the work will last.
Database 8. Volumes of resources (BA_08). Reactions are the amount of resources.
Database 9. Units of measurement (BA_09). In Table U_R are units of measurement of resources.
Database 10. Relationships between works (BA_33). A reaction is a type of communication
between robots.
Database 11. Robot resources (BA_34). Table U_R – shows the ratio of resources to works.
Database 12. Lag / advance of related works (BA_10). It is used to determine the delay/advance of
the subsequent work in relation to the preceding work.
In addition, internal tables of project plans were created in the RISPP environment (outside of the
given databases), which were not related to each other. They were used to display the information
formed during the work of the RISPP in the environment of instrumental software tools (including
MS Project ):
1. Projects (_Mod06).
2. Project work (_Mod03).
3. Letter of resources (_Mod04).
4. Handbook of time measurements (_Mod07).
5. Directory of units of measurement (_Mod09).
6. Handbook describing the links between the works (_Mod10).
7. Relationships between works (_Mod33).
8. Resources for robots (_Mod34).
The use of tables of intellectual modules for developing reflexes and developing reactions of each
base was determined by the definition of the content and meaningful scenario of the action, which is
the essence of the input data flow. A method was proposed and an algorithm for controlling the
process of content identification, content scenario and RISPP reaction was developed. The method
allows it to gradually determine the content of the input data stream in relation to the actions to be
performed in the project planning process. Main stages:
1. Obtaining the input data stream.
2. Determining the content of the input data stream (database BA_00).
3. Defining a meaningful scenario of the input data stream (database BA_01).
4. If the meaningful action scenario does not correspond to the content - clarification from the
operator. Return to item 1.
5. Step-by-step processing of the meaningful action scenario.
6. Selection of the next element of the meaningful action scenario.
7. Defining a fragment of text related to the next element of the meaningful scenario of actions
(base BA_02, base of directories: BA_03, BA_04, BA_06).
8. If the response of RISPP is a command to planning - the execution of the command in the
planning management module.
9. If the reaction of RISPP is an element of the project plan - filling in the relevant table of the
project plan.
10. If the meaningful scenario is not fully processed - then transition to paragraph 6.
11. End of work.
The system activity diagram is shown in Fig. 3.
A demonstration version of the reflective intelligent project planning system has been developed,
which is configured for the input data stream in the form of text in human language, which can come
from all sources specified in paragraph 2.3 - from members of project management teams,
information standard and documentation. This system performs the following functions:
Figure 3: UML activity diagram RISPP
1. Processes arbitrary input data streams.
2. Learns (with the help of an instructor) to respond properly to incoming data streams.
3. Implements actions for preparation of PSND (forms tables of project works, communications,
resources).
4. Memorizes the context. If the incoming data stream determines a project or node of work,
then in the future the system is working with this project or node.
5. Generates various reports.
Experimental studies were conducted using a demonstration version of RISPP [10]. The research
was conducted with PSND of 2 different development projects. The uniqueness of works in these
projects amounted to more than 90%, and the uniqueness of resources - to 100%. The project plans
contained 1,221 jobs and about 1,000 connections. Resources were used in 919 works. It was assumed
that this information could come from project management team managers, from the information
standards, and from the documentation. To make the system more reliable, the input data streams
were distorted by replacing or removing 10% and 20% of the letters. The input data streams (more
than 3000) contained information on project creation, tasks and resources, establishing links between
works, allocating resources by works and setting numerical parameters: duration of works, amount of
resources, lag / advance of works. Experimental studies have shown the high efficiency of
recognizing the content of human language appeals. The probability of correct definition of the
content for undistorted texts was about 97.42%, and the content scenario reached 92.68%. For the
distorted it was: for 10% distortion - the content of 96.80%, the meaningful scenario - 88.89%; for
20% distortion - content 95.31%, meaningful scenario - 86.80%. The implementation of the module
for cancellation of incorrect actions (cancellation/incorrect/rollback/deviations) in RISPP gave a result
that allows us to proceed to the creation of an industrial reflexive intelligent project planning system.
3. Results
Methods, algorithms and demonstration tools of the reflex intelligent system of project planning
are developed. It is proposed to divide the structure of the information database by the type of
reactions into 12 databases. Each database is used to generate responses in terms of planning
management and the formation of PSND to calculate the project plan. PSND formation is performed
in sections: WBS; project works; resources and their volumes required to carry out project work;
connections between works. Algorithms and software tools for teaching RISPP on educational data
have been developed. In the process of learning in RISPP reflexes are developed to determine the
reactions that are adequate to the content of the information coming into the incoming data streams. In
the project planning process, these reflexes are used to determine the actions of the intelligent system
for the formation of PSND, which is used to calculate the project plan. This made it possible to
achieve the goal of the research - to develop a reflexive intelligent system of project planning, which
is capable of forming the PSND of the project with minimal human participation. The system was
tested in experimental studies. The results of experimental research have shown that the developed
system allows to quickly and efficiently build PSND based on the use of standard information and
verbal interaction with the project team. At the same time, human participation in solving this creative
task is limited due to the increased intelligence of the project planning system. This, in the long run
will lead to the development of more realistic plans, and thus increase the likelihood of clear work of
managers and executors to implement projects "according to plan".
4. Discussion
These results made it possible to achieve the goal of the research, namely to develop a reflexive
intelligent system of project planning, which is capable of forming the PSND of the project with
minimal human participation.
The limitations of this study are that the system only generates a Project Schedule Network
Diagram for use in the CPM method. And the functions of the calculation itself, as well as the use of
other methods, in particular the PERT method, are not considered. Although based on the fact that the
system is based on statistics of previous projects, the use of the PERT method to determine the
optimistic, pessimistic and probable deadlines for project completion would be appropriate.
The point of discussion is the use of a reflex approach to create an intelligent planning system,
rather than, say, neural network technologies. First, it involves the use of different types of data that
influence project decisions (verbal and formal). Secondly, the use of neural networks involves
significant costs for building and training a network capable of recognizing some reaction in the
incoming data stream. Especially if the input data stream is divided into elements. So, for example, in
the course of experiments 135,857 elements were allocated. And despite such a large number of
elements on which reflexes are made, the reaction of RISPP to the incoming data stream was almost
instantaneous (less than 1 second). In this case, when new information appears, such a neural network
must be retrained. This problem is absent in reflex systems. They choose reactions in accordance with
developed reflexes, which are based on statistics.
The disadvantage of the study is that the methods of building expert systems, in particular, the
knowledge base, are not used. This could significantly expand the capabilities of such a system.
But as seen from the results of experimental research - the system is simple and effective and can
be used in practice by project managers responsible for planning.
5. Conclusion
Research has been conducted on the development of a reflective intelligent project planning
system that would be able to form a PSND project with minimal human involvement, which therefore
requires the use of:
project management methods, to determine the data that will be processed in the system and
to determine the set of actions required in the process of creating a PSND and calculating the
project plan;
information technology methods - to collect information to develop a project plan;
statistical methods and methods of artificial intelligence - to create an intelligent project
planning system;
methods of computational linguistics - for processing human language information.
As a procedural core of the system, a modified reflex method was used, with adequate actions for
the formation of the project plan on the basis of information, expertise and documentation
accumulated in previous projects.
The new scientific result is digital project management technologies through the use of a reflex
approach and a method of forming PSND with minimal human participation. This work makes a
significant contribution to the creation of digital project management. The obtained result
intellectualizes the project planning process, which reduces the cost of creating a project plan and
increases its accuracy and realism. Which in turn will increase the likelihood of timely
implementation of the project plan.
References
[1] E. Głodziński, M. Szymborski, Utilization of software supporting project management in middle
and large project-based organizations: an empirical study in Poland, Communications of the PCS
164 (2019) 389–396. doi:10.1016/j.procs.2019.12.198.
[2] A. Lester, 51- Primavera P2, in: Managing Engineering, Construction and Manufacturing
Projects to PMI, APM and BSI Standards, Project Management Planning and Control,
Butterworth-Heinemann, 2021, pp.485–503. doi:10.1016/B978-0-12-824339-8.00051-1.
[3] M. A. AbdEl-Migid, D. Cai, T. Niven, J. Vo, K. Madampe, J. Grundy, R. Hoda, Emotimonitor:
A Trello power-up to capture and monitor emotions of Agile teams, Communications of the SS
186 (2022). doi:10.1016/j.jss.2021.111206.
[4] V. Morozov, O. Mezentseva, Development of Optimization Models of Complex
Infocommunication Projects Based on Data Mining, in: IEEE International Conference on Smart
Information Systems and Technologies (SIST ‘21), Nur-Sultan, 2021. pp. 1–7.
doi: 10.1109/SIST50301.2021.9465991.
[5] Y. Kovtunenko, The application of artificial intellectums in the enterprise management system:
problems and advantages, Communications of the eg OPU 2 (2019) 93–99. doi:
10.5281/zenodo.4171114.
[6] A. Berko, O. Yavlinskyi, Intellectual support system makes decisions when managing non-
revenue projects, Communications of the LPNU 653 (2009): 12–23. udc: 004.89
[7] R. Serbul, Information and analytical management system of the project team in software
development, Master’s thesis, International Institute for Applied Systems Analysis (IIASA),
Kyiv, Ukraine, 2017.
[8] A. Srarukh, Application of expert systems in business environment, Communications of the sb
IHU 41 (2020) 114-121. doi:10.32841/2413-2675/2019-41-15
[9] I. Teslia, V. Pylypenko, N. Popovych, O. Chornyy, The Non-Force Interaction Theory for Reflex
System Creation with Application to TV Voice Control, in: Proceedings of the 6th International
Conference on Agents and Artificial Intelligence, ICAART ’14, Leria, 2014, pp. 288–296. doi:
10.5220/0004754702880296.
[10] I. Teslia, N. Yehorchenkova, I. Khlevna, O. Yehorchenkov, Y. Kataieva, G. Klevanna,
Development of reflex technology of action identification in project planning systems, in:
Proceedings of the International Conference on Smart Information Systems and Technologies,
SIST ‘22, Nur-Sultan, 2022. pp. 269–274.