See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/348327272

Project Management using Machine Learning

Technical Report · November 2020

CITATIONS READS

0 5,616

2 authors:

Ishaan Buch Kirtan Pokiya

Nirma University University of Denver
9 PUBLICATIONS 0 CITATIONS 4 PUBLICATIONS 4 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Ishaan Buch on 31 May 2021.

The user has requested enhancement of the downloaded file.

 Project Management using Machine Learning

Ishaan Buch Kirtan Pokiya

18BCE079, CSE B 18BCE095, CSE B
Institute of Technology, Nirma Institute of Technology, Nirma
University University
Ahmedabad, India Ahmedabad, India
[email protected] [email protected]

Abstract—Any project cannot be implemented and III. TRADITIONAL PROJECT MANAGEMENT

expected to be a success without a proper planning process.
Project management is a collection of activities which help Traditional project management approach works for
organising a project and provides a path for it to be successful. many domains like a brand spanking
Machine Learning is one of the upcoming technologies which is new product’s development. In ancient project
helping in simplifying and making. Some work processes more management, the whole project life cycle is
and more efficient with its techniques. Machine Learning can split into parts and therefore the work is completed by
be applied to project management in order to find the optimal phases. Tools like Work Breakdown Structure, Gantt
way to go for project development.We will see how machine Charts are extensively utilised in ancient project
learning benefits and how it can improve the process of project management.
management.
IV. PROJECT PLANNING
Keywords—project management, machine learning For most projects, it's potential was depended on the
team to outline and establish the whole demand of the
I. INTRODUCTION u l t i m a t e p r o d u c t d i r e c t l y. T h e r e f o r e t h e
Project management is the process of applying skill, whole project was planned. In such circumstances, they are
methods, processes, knowledge and experience to achieve a do not anticipate several changes. the whole project
goal set for the development of a software project. Project development is foreseeable. A series of
management has a final goal which has to be achieved and activities are designed for the project like initiation,
certain activities and process are designed in order to achieve planning, execution, watching and dominant, and
it. Machine Learning is one of the most used domains used in shutting etc. All the activities are carried out in a
recent times, it mainly focuses providing the system the certain sequence. Ancient project management usually
ability to learn and improve without explicitly programming used waterfall model of development, where a part is to be
it. Project management is a process followed from years and completed before starting the next.
the use machine learning in different stages of project
management have been highlighted ahead. While in another reasonably project like new software
system application or perhaps a futurist new development, it
II. PROJECT MANAGEMENT is troublesome to outline all the requirements at the initial
stage. Such projects cannot use the traditional methods of
Project management is a methodology which has been project management and need a more versatile approach
used since long. It follows a hard and fast sequence of towards project management like Agile, spiral etc..
initiation, planning, execution, monitoring, and closure.
Project management is an ancient practice, it has been A. Phases of normal Project Planning approach
applied to various projects like the Taj Mahal and the 1. Initiation
Pyramids.
2. Planning
Project Management is a collection of well-
tried practices, tools and techniques that are applied in 3. Execution
managing a project. A project could be a temporary 4. Monitoring
endeavour undertaken to form a product, service or result.
5. Closure
Project management is a process involving certain
activities. They are: V. COST ESTIMATION
• Project Planning and Tracking Cost estimation is a technique to find estimates of cost
which is the basic financial spend on the efforts to develop
• Project Risk Management the project. Cost estimation models are basically
• Estimation Management mathematical or parametric equations which are used to
estimate costs.
• Scope Management
A. COCOMO MODEL
• Configuration Management
Constructive Cost Model or the COCOMO Model is one
• Project Communication Management of the widely used model in cost estimation for software
• Configuration Management products. It is procedural cost estimation model proposed by
Barry Boehm in 1970. The COCOMO Model has had
• Project Resource Management various changes and has been introduced with different
versions of it.
The two key parameters that define the quality of a software
product are:
 1. Effort Basic COCOMO assumes that effort is a function of only
number of lines of code and nothing else. There are various
2. Schedule
Different COCOMO Models have been introduced to COST VERY
LOW LOW NOMI
NAL HIGH VERY
HIGH
predict cost estimations at different levels of a software DRIVERS
project. Boehm has classified the software projects into three
main categories to help calculate costs easily. They are: Product
Attributes
1. Organic: It is a small software project where the team
size is small and the problem statement has been well
understood or has been previously solved. Required 0.75 0.88 1.00 1.15 1.40
Software
2. Semi-detached: This is a software project which lies Reliability
between organic and embedded software projects. This Size of 0.94 1.00 1.08 1.16
requires a larger team size compared to organic projects. Application
And the problem statement is bit tricky to comprehend Database
and requires better guidance.
Complexity 0.70 0.85 1.00 1.15 1.30
3. Embedded: This kind of software project requires the of The
highest level of complexity, creativity, and experience. Product
The team size is considerably larger than the other two Hardware
types of projects and the teams require much more Attributes
experience.
All the types of projects acquire different value of Runtime 1.00 1.11 1.30
constants. Performance
Constraints
The COCOMO Model is consists of three different
Memory 1.00 1.06 1.21
models based on hierarchy. Any of the three can be used Constraints
based on our requirements. They are :
1. Basic COCOMO Model Volatility of 0.87 1.00 1.15 1.30
the virtual
2. Intermediate COCOMO Model machine
3. Detailed COCOMO Model environment
Required 0.94 1.00 1.07 1.15
1. Basic COCOMO Model turnabout
time
The Basic COCOMO Model can be used for quick and
rough calculations of Software Costs. Personnel
attributes

Analyst 1.46 1.19 1.00 0.86 0.71

capability

Applications 1.29 1.13 1.00 0.91 0.82

experience

Software 1.42 1.17 1.00 0.86 0.70

engineer
capability
Fig 1: Equations for calculation of Basic COCOMO Model
Virtual 1.21 1.10 1.00 0.90
machine
The following are the values of all the constants as per experience
the classification of different type of projects:
Programming 1.14 1.07 1.00 0.95
language
SOFTWARE A B C D experience
PROJECTS
Project
Attributes
Organic 2.4 1.05 2.5 0.38

Application 1.24 1.10 1.00 0.91 0.82

Semi 3.0 1.12 2.5 0.35 of software
Detached engineering
methods
Use of 1.24 1.10 1.00 0.91 0.83
Embedded 3.6 1.20 2.5 0.32
software tools

Table 1: This table shows the values of the constants used for calculations
Required 1.23 1.08 1.00 1.04 1.10
as per the classification of projects
development
schedule
2. Intermediate COCOMO Model
other factors which are involved such as reliability,
Table 2: This table shows the cost drive factors for the Intermediate experience, capability etc. These factors are known as Cost
COCOMO Model
Drivers.
 The project manager has to scale these parameters on the main types of learning algorithms are used. They are instance
scale of 1-5 with 3 being normal. Depending on the scale based learning and regression algorithms.
given by the appropriate values are taken from the above
table and multiplied to calculate EAF (Effort Adjustment A. Instance Based Algorithms
Factor). The intermediate formula for effort is: In machine learning, instance based learning rather than
generalising the problem, it predicts outcomes of the dataset
Effort =(a(KLOC)b)*EAF based on past training from different datasets. In simple
The values of a and b of the intermediate model are words, it learns from past datasets and does not generate a
generalised equation for the variables but it compares with
the past datasets for the new dataset and gives suitable
SOFTWARE PROJECTS A B
outcomes. This type of model is built from a sample
database, and compares them to the new dataset and makes
Organic 3.2 1.05 prediction based on similarities between them. It is also
known as memory based learning. This is one of the simplest
Semi Detached 3.0 1.12 algorithms to implement in as it does not require to create a
general solution for the new instances whereas it just goes
through the stored database in order to make a prediction.
Embedded 2.8 1.20
The biggest disadvantage of these algorithms is the space
required by it. As it stores all the past instances to make a
B. Detailed COCOMO Model prediction, it requires large storage space to store the sample
Detailed COCOMO includes all the attributes of the database. Another disadvantage of this algorithm is the time
intermediate COCOMO model as an assessment for each required to make a prediction. As the prediction is based on
step of the software development process. In detailed all the past instances, the model will traverse through the
COCOMO the whole project is divided into different whole database which will require significantly more time
modules and the effort calculation is done for each in the than a generalised model. Some examples of instance based
same way as intermediate COCOMO and then sum the learning are K-proximity algorithm (KNN), Self organising
effort. Map (SOM), Learning Vector Quantisation (LVQ), Locally
Weighted Learning (LWL), etc.
The six phases of detailed COCOMO are:
B. Regression Algorithms
1. Planning and requirements Regression analysis models are statistical models which
2. System design form relationships between the dependent and independent
variables or predictors. These models are usually used to
3. Detailed design predict time series or find relationship between two
4. Module code and test variables. Regression analysis is one of the most commonly
used statistical model in machine learning. Some examples
5. Integration and test of regression models are Ordinary Least Square Regression
6. Cost constructive model (OLSR), Linear Regression, Polynomial Regression,
Logistic Regression etc.
The efforts of each process are calculated and the sum is
then used for further calculation ADVANTAGES OF MACHINE LEARNING IN SOFTWARE
PROJECT MANAGEMENT
VI. RISK MANAGEMENT The following are the advantages of using machine
Risk management is a collection of processes which learning in software project management:
include identifying, addressing and eliminating the problems
which might harm the cost, schedule or technical success of 1. Reduced Expenses: According to a study
the software product. by MIT Sloan, 63% of the organisations
eye website price reduction as their main motivation
Risk management involves two major activities, they are for adopting Machine Learning technologies.
risk assessment and risk control. Risk assessment involves Machine Learning will facilitate automatise repetitive
identification and analysis of the potential risks that could manual tasks, permitting groups and project managers
hamper the progress or the final product. After identifying all to specialise in additional high worth delivering
the potential risks, the next step is manage them so they do activities, thereby reducing
not affect the progress of the project. labor prices whereas rising work quality.
VIII. FLAWS OF TRADITIONAL PROJECT MANAGEMENT 2. Machine Learning -Based Analytics: Prognostic analytics
is another nice profit that AI provides. The
Traditional project management system will gather past project info and forecast, with
approach depends heavily on up-front thinking. high degrees of accuracy once comes are going to
The principle is that if you propose, execute the process, and be completed. It may also give managers and team
take corrective action then the project is going to members with additional visibility by alerting them if the
be prosperous. The sequence of project goes off-course.
events would perpetually needed to be reformulated
as queries concerning scope and technology are resolved. If 3. Higher Efficiency: clasp Machine Learning in project
the developer is unaware of all the challenges ahead then it management whenever potential can result
would lead to the failure or delay to deliver the software in higher results thanks to inflated accuracy and
product. better potency. Machine Learning has the power to offset
potential risks victimization knowledge points and
IX. MACHINE LEARNING may crunch that knowledge in time period as a project
There are various different learning algorithms in progresses – abundant quicker and additional accurately
machine learning so many different algorithms are used to than most humans.
improve the efficiency of project management. Mainly two
X. APPLICATION OF MACHINE LEARNING IN SOFTWARE PROJECT XI. CONCLUSION
MANAGEMENT With all the benefits of machine learning, it’s a no
ML can be used to bring several changes in software brainer for project managers to be told concerning and adopt
development. machine learning technologies. The business is shifting to
embrace machine learning and it’s progressively being added
A.Detecting code deviations to project management as a core part of the programme.
Software developers of companies write millions of lines
of codes for their software product. The main aim is to assure XII. REFERENCES
that the code does not deviate from the actual aim because 1. Miguel Pedroso, May 2017, “Application of Machine
any deviation will lead to result in change of code quality. Learning Techniques in Project Management Tools”
The quality of code is no more a simple defect but it is now a
proactive assurance of the workability and efficiency of the 2. Somya Goyal and Anubha Parashar, November 2017,
software project. A machine learning tool can be constructed “Machine LearningApplication to Improve COCOMO
for code analysis where developers can use few examples of Model using Neural Networks”
source code guidelines to train a decision tree classifier to 3. GeekForGeeks, Software Engineering | COCOMO
find code violations in their codes. Model,”https://www.geeksforgeeks.org/software-
B. Cost Estimation Model engineering-cocomo-model/?ref=lbp”
Many variations have been made to the COCOMO 4. Miroslaw Ochodek, November 2019, “Recognising lines
Model using machine learning in the past years with the of code violating company-specific coding guidelines
latest being ANN using dragonfly algorithm. The software using machine learning”
process improvement is facing so many challenges like lack
of data for analysis and lack of information regarding data 5. KnowledgeHut, Traditional Project Management,”https://
types. The data quality problem is pretty prominent as it does www.knowledgehut.com/tutorials/project-management/
not only relate to consistency but also the quality of traditional-project-management”
completeness of the dataset. The is no effective technique 6. B. Depaire, December 2019, “https://bookdown.org/
using machine learning used for software cost estimation but content/1785/“
the research work is continued.
C. Risk Management
To develop a machine learning algorithm which makes
risk prompts requires a set of distributed machine learning
tasks. As each of the risk prompts are independent of each
other so their taxonomies are independent of each other.
These simplifications reduce the problem significantly, and
we have several multi-label classification problems, which
are resolved by using multiple multi-label artificial neural
networks (ANNs).

View publication stats