Predicting the Progress of Vehicle Development Projects:

An Approach for the Identification of Input Features

Oliver Böhme

and Tobias Meisen

Chair for Technologies and Management of Digital Transformation, Bergische Universität Wuppertal,

Rainer-Gruenter-Str. 21, Wuppertal, Germany

Department of Electrical Engineering, Information Technology and Media Technology, Bergische Universität Wuppertal,

Rainer-Gruenter-Str. 21, Wuppertal, Germany

Keywords: Machine Learning, Input Features, Automotive, R&D, Project Success Indicators, Critical Success Factors,

Employee Survey.

Abstract: Today project managers estimate time and other project relevant key performance indicators by using project

management tools e.g. milestone trend analysis. We believe that predicting the project’s progress with

traditional methods will soon reach its limitations due to the increasing complexity in vehicle development.

Machine learning methods provide one possible solution. The vision is to predict the progress of development

projects in the early stages of the project. In order to make this vision come true, we need to define measurable

input features for machine learning models. In this paper, we focus on representing an approach to identify

parameters that exert influence on the progress of development projects.

1 INTRODUCTION

For many customers, the quality of a product is one

of the key factors in a purchase decision. As a result,

many brands within a market try to differentiate

themselves through this purchase criteria. In addition

to the constantly growing demands from the

competition and from the consumers and the

government (Deloitte, 2019), the challenges in the

development of innovative products also grow with

the advancing technological progress. By now more

than 90 percent of all future vehicle innovations are

located in the field of electrical/ electronics

development, the search for errors and the sustainable

correction of them contribute greatly to the success of

the project.

With the considerable increase in the functional

integration of the electronic control units (ECU), the

complexity of product quality management increases

considerably, since the system behaviour is no longer

deterministic. This refers in particular to the

interaction of the ECUs in the overall

interconnection. Each component is known exactly,

but a prediction of its state of integration cannot be

made with great certainty. For example, a new

software delivery of a highly interconnected ECU can

result in a new function in the vehicle, while at the

same time some other functions are no longer

available due to regression in the software.

Furthermore, the products may be developed in better

quality, at lower costs and in less time. The

requirements for error elimination and minimization

are increasing. At the same time the demands on

meeting deadlines are increasing, as milestone shifts

often result in considerable additional work and costs,

an extension of the project duration or in a reduction

in the scope of services. From our experience, as

industrial engineers working in a German car

manufacturer, we have accompanied the course of a

large number of vehicle projects and derivatives from

the perspective of electrical/ electronics development.

In many cases we were able to observe that a

prediction of the project duration with traditional

methods could only be made insufficiently or with

great uncertainty. Depending on the project’s

complexity either temporal deviations in the function

development or in the reduction of errors may occur,

because the courses of past vehicle projects were not

sufficiently considered.

Nowadays on the other hand, we can already see

numerous applications from the field of artificial

intelligence (AI) where historical data is used to

predict future outcomes, e.g. sales forecasting,

522

Böhme, O. and Meisen, T.

Predicting the Progress of Vehicle Development Projects: An Approach for the Identiﬁcation of Input Features.

DOI: 10.5220/0010187905220530

In Proceedings of the 13th International Conference on Agents and Artiﬁcial Intelligence (ICAART 2021) - Volume 2, pages 522-530

ISBN: 978-989-758-484-8

predicting stock prices or churn analysis. Many of

these applications are already on an almost human

expert level, e.g. early diagnosis of diseases based on

X-ray images or AI-based translation of languages

(Kermany et al., 2018); (Hassan et al., 2018). This

leads to the question of whether the usage of methods

from the field of AI can also be used for the prediction

of the project’s progress, which is exactly our

scientific vision. Having a system like this, the risks

of shifting elementary project milestones in the early

phase could be recognized in order to create the

preconditions for steering the project into an orderly

path again. This ambitious goal requires not only the

definition of project’s progress, but also the

identification of indicators and factors that exert

influence on the progress of development projects.

With this contribution we intend to identify indicators

that are suitable for the use in machine learning

scenarios. This approach will be presented on the use

case of the electrical/ electronics development of a car

manufacturer.

For this purpose, Section 2 provides an overview of

project success models and critical success factors as

a result of a broad literature research in order to

identify possible general starting points for input

features. Section 3 presents the research method, an

employee survey on the topic of project success in the

electrical/ electronics development department. With

a precise view on the research area, the starting points

identified in section 2 will help to extract the domain

knowhow and to find possible input features for the

use case. Section 4 compares the results from the

literature research with those from the employee

survey. With regard to applicable input features for

machine learning models, a critical look is taken at

the analysis results. A summary and further

proceedings are presented in section 5. An overview

of the identified input features is given. Finally, a

recommendation for further work is given and as a

scientific vision the concept of our machine learning

approach is presented.

2 RELATED WORK

Based on the vision of being able to predict the course

of vehicle development projects, we first need to

identify the right measurable factors influencing the

course of the project. Therefore, in this section we

will take a look at literature for applied machine

learning approaches in project management and

project management literature regarding the concept

of project success as well as critical success

indicators.

2.1 Applied Machine Learning

Looking at applied machine learning the number of

approaches has been increased greatly during the last

20 years. By now there are also various applications

for machine learning in project management that have

been discussed in literature. Popular approaches are

for example: estimation of software development

effort (Srinivasam and Fisher, 1995), early life cycle

cost estimation (Boetticher, 2001), machine learning

in scheduling (Aytug et al., 1994), software project

risk management modelling (Hu et al., 2007),

predicting the priority of reported bugs (Sharma et al.,

2012) and software requirements prioritization

(Perini et al., 2013). Interestingly our literature

research did not reveal the use of any machine

learning algorithms applied in the early life cycle for

the estimation of development project’s progress in

the automotive sector. Since the complexity of

product development - driven by increased

requirements due to intensified competition, regional

regulatory requirements and market-specific

customer wishes - is considered to be higher than in

projects from other industrial sectors, the automotive

industry demands a tailor-made approach. Though the

available literature focuses on very specific

approaches for time estimation in projects for various

industrial sectors except automotive.

Huang and Chen developed a framework for

estimating the project completion time. Their

framework starts with collecting data such as project

task structure, task relations, and quantified team

member characteristics in order to analyse the

influence of these factors on the project completion

time. Further a simulation model is used to assign the

identified tasks dynamically to the team members

according to their knowledge level and other factors.

After several iterations the final value of time is

estimated. Finally, they analysed the data to identify

significant factors influencing the project completion

time (Huang and Chen, 2006). While this framework

also deals with the analysis of influencing factors on

the project duration, the approach does not offer an

opportunity for adaptation into a machine learning

approach. Accordingly, the proposed method of

Huang and Chen is more suitable for low complexity

projects and therefore not suited in the scope of this

paper.

Another approach was published by Pedroso. He

proposed a system that aimed to help improving the

planning process in project management by

performing risk analysis. Therefor instance-based

learning and regression models were used, which

gave satisfiable results when applied on real-world

Predicting the Progress of Vehicle Development Projects: An Approach for the Identiﬁcation of Input Features

523

scenarios (Pedroso, 2017). With this Pedroso has

developed an interesting approach to predict risks for

future projects based on the work history of a specific

project manager. However, a transferability to the

context of vehicle development can be doubted as the

approach lacks the ability to capture global patterns.

It is assumed that only the identification of cross-

project features will allow a prediction of project's

progression in the required quality.

Li et al. developed a method for time estimation in

ship block manufacturing. By using the k-Means

algorithm the researchers clustered the different ship

blocks according to their features (e.g. length, width,

depth, weight, form etc.). Afterwards and in order to

evaluate the planned time of each cluster, Li et al.

used a data envelopment analysis model. By

processing the calculated results an estimation for the

manufacturing planning was accomplished. Thereby,

they used a genetic backpropagation neural network

to capture the knowledge for reuses (Li et al., 2019).

With their approach, Li et al. were able to predict the

manufacturing time of a ship by forecasting the

respective times for the production of a single ship

block and then adding these times considering the

number per block type. Unfortunately, the

identification of the relevant features has not been

described in detail by the authors. Thus, this approach

is strongly industry- and business-specific. It can only

be transferred to a very limited extent to another

business area (development) of another industry

(automotive engineering). Finally, the approach does

not consider the high complexity of automotive

product development, resulting primarily from

interconnectivity and the collaboration of ECUs on

different vehicle architectures.

Ahmed et al. discussed a method for the estimation of

the procurement time of Public-Private-Partnerships

projects. By using multiple regression models, they

compared the predicted procurement time with

secondary data from the World Bank (Ahmed et al.,

2019). With regard to the identification of factors that

can cause a delay in PPP projects, Ahmed et al.

conduct a literature search and evaluate three case

studies. As with the other approaches above, these are

industry-specific factors. In addition, the scope of the

referenced projects is only set up to the procurement

of resources. The true complexity in the temporal

forecast of projects begins however, due to its

characteristic uniqueness, with the operational start of

the project.

2.2 Defining Project Success

The concept of success in project management and its

composition are topics that have in principle been of

interest to the research community - and naturally

also to practice - since the beginning of the

development of project management as an

independent discipline (Jugdev and Müller, 2005).

Not without reason it is called the most discussed

topic in the world of project management (Shenhar et

al., 1997).

The analysis of various papers shows that many

researchers have common intersections. In the

research period between 1960 and 1980, researchers

concentrated mainly on the “magic triangle” (cost,

time and quality) (Atkinson, 1999). In the period from

1980 to the turn of the millennium these factors retain

their essential role in measuring project success and

are simultaneously extended to include stakeholder

satisfaction (e.g. end customer, end user, etc.) as well

as the benefits for stakeholders and the supporting

organization resulting from the project. This research

direction is now accompanied by an awareness

between the project implementation phase and the

success of the project after the end of the project

(Pinto and Slevin, 1988) (Wuellner, 1990) (Pinto and

Pinto, 1991). From between 1990 and the 21st.

century onwards, strategic goals and overall business

success increasingly play a decisive role as the

validity of the research continues and the success

factors determined up to this point are taken as a

basis. Projects are also increasingly being evaluated

on the basis of specific definitions of success and

failure related to the individual project (Navarre and

Schaan, 1990) (Shenhar et al., 1997) (Baccarini,

1999). It should also be mentioned that, depending on

the domain, additional criteria such as safety or

environmental friendliness are also included (Kometa

et al., 1995) (Kumaraswamy and Thorpe, 1996).

Different researchers have identified different success

criteria. One possible cause may lie in the different

domains themselves. Just as the requirements for

products from different industries vary, the domain-

specific project success criteria could also be

divergent.

2.3 Critical Success Factors in Projects

In contrast to project success indicators, critical

success factors (CSFs) are all those activities that are

intended to ensure the success of organizations and

projects when properly implemented (Boynton and

Zmud, 1984). Müller and Jugdev also define them as

those elements and independent variables which,

when influenced, increase the probability of success

of projects (Müller and Jugdev, 2012).

ICAART 2021 - 13th International Conference on Agents and Artiﬁcial Intelligence

524

Since the beginning of the activities in this research

field, a large number of researchers in various

industrial sectors and countries have been dealing

with critical success factors in projects (CSF). A

comparison of the various publications shows that a

certain set of CSFs is repeatedly highlighted.

In the research period up to 1980, the researchers

identified a precise project definition with clearly

defined objectives, detailed and realistic project

planning and effective project monitoring and

controlling as the most important CSFs. Furthermore,

the technical and leadership skills of the project

manager and the project team, effective

communication during the project and support from

top management are the most important factors (Pinto

and Slevin, 1986) (Pinto and Slevin, 1988) (Belassi

and Tukel, 1996).

The evaluation of the following years up to the turn

of the millennium confirms the results of the analysis.

The technical and leadership competencies and the

support of the management board rise in the ranking

list of the most frequently mentioned CSFs to the

ranks one and two. The following ranks reflect the

factors from previous years. But also, the

understanding for the needs of the project customer

e.g. attained by the active integration of the product’s

user seems to become ever more relevant for

enterprises of all sectors (Fortune and White, 2006).

The above impression can also be confirmed by

looking at the research period up to 2010. The CSFs

identified over the past ten years could be repeatedly

identified during this period in a slightly changed

ranking order. A look at the following years

repeatedly confirms the research results from the

beginning of the observations. Those who manage

their projects with clear goals, precise planning and

effective monitoring, risk and change management

obviously have good prospects for project success. If

this skill set can now be paired with technical and

leadership skills in the project team, effective

communication and sufficient resources, the

probability of success can be increased again. In

addition to a good understanding of the customer’s

needs, for the first-time researchers are also counting

positive relations to politics and society among the

most frequently mentioned critical success factors.

3 RESEARCH METHOD

After having looked at the project management

literature, we now carry out a precise analysis of the

application case. In order to extract the domain

specific knowhow in the related research area we

designed a questionnaire to conduct an employee

survey. For this survey a group of 80 participants

were asked on different aspects of time-related and

process-oriented factors and about widely known

critical project factors in their individual working

environment. To make sure of getting knowledge

from experts, only participants with at least ten years

of working experience in electrical/ electronics

development were chosen. In addition, attention was

paid to a heterogeneous group composition (e.g. age

between 30 and 60+). These experts were selected on

the basis of their individual role (e.g. executives,

project managers, test engineers), their project

affiliation (e.g. small size projects, medium size

projects, full size projects) and their responsibility

within the specific project (either development or test

and integration), all to make sure, that we will receive

every possible perspective on the project’s pasts.

For this questionnaire, we identified 17 items to test

the research area for the presence of critical success

factors. This includes a strict project planning, the

consequences of non-compliance with milestones and

processes in the early project stages or a delay in

assigning suppliers. Late concept decisions, the

management of risks and the support of top

management are also examined. It is also analysed

whether the scope of services, the technical

specifications, the employee satisfaction or a stable

internal company policy contributes to the success of

the project. For each statement in the questionnaire

we asked the participants on a scale from 1 (“does not

apply at all”) to 5 (“is absolutely true”), how

applicable this statement is. In order to avoid false

statements as far as possible, the participants were

also given the opportunity to choose "don't know" or

"not specified".

4 FACTORS INFLUENCING THE

PROJECT’S PROGRESSION

The analysis of the relevant literature on applied

machine learning from section 2.1 has shown that

there are only few and very specific applications in

project management at the moment. This shows the

great potential that the applied machine learning has

in the field of project management. At the same time,

this also shows a large research gap.

4.1 Examination of Project Success

Indicators

Since the problem of predicting the progress of

Predicting the Progress of Vehicle Development Projects: An Approach for the Identiﬁcation of Input Features

525

vehicle development projects is a highly

interdisciplinary one, the relevant project

management literature between the 1980s and today

was examined with regard to the concept of project

success and critical success factors.

From this we can see that there is a lack of a universal

definition of project success. This leads to the need

for a working definition on the basis of which the

continuation of further research activities can be built.

For this purpose, the project success indicators

identified in section 2.2 are combined into clusters

and then analysed for their relative frequency (figure

1). The results show that the classical success

indicators such as costs, time, quality and scope of

services can be counted among the most frequently

cited. Furthermore, it can be read from the results that

the right management of stakeholders (customers,

clients and project managers and employees) can be

an indicator for a successful project. However,

project-specific topics such as effectiveness in

implementation and technical specifications are also

regarded as indicators of success by 40 to 50 percent

of researchers. Restricting the focus of the study to

the period from 2010 onwards confirms the above

evaluation in most of the citations. Factors that have

been confirmed since then are revenue, profit and

strategic factors such as the development of new

market shares or new markets, further the

development of the organization or the increase in

competitiveness can now be found with a high

relative frequency. After it used to seem sufficient to

complete projects in the right time, at the right costs,

in the right scope and with the right quality, now a

measurable “return on investment” is increasingly

coming to the fore.

Figure 1: Top cited project success indicators ranked by

relative frequency.

Based on the framework of this work and the

scientific question at the core, only those indicators

will be considered for further activities that can

generally measure the success within the project

implementation. Since indicators such as customer

satisfaction and client satisfaction, revenue or profit

and the possible development of new market shares

can only be measured during or after the project

handover or within the context of the use of the

resulting product, these indicators should not be

considered here further. Similarly, the satisfaction of

the project managers shall be excluded, since they

define the project success by the fulfilment of the

defined project goals and their satisfaction can

therefore only be measured at the end of the project.

Furthermore, the costs as an indicator for the success

of the project shall not be further considered, since

these are triggered indirectly via budget and resource

availability in the research area. From this limitation

a working definition for the project success (figure 2)

can be deduced. With regard to our vision, the

prediction of the project’s progress, the quality, the

scope of services, the technical specifications, the

effectiveness/ efficiency in project implementation

and the satisfaction of the project team members shall

be defined as primary success indicators.

Figure 2: Working definition project success indicators.

As described in section 1, the success of a vehicle

development project is to be measured from the

perspective of the electrical/ electronics development

department. For this concrete use case, the project

success therefore requires a defined scope of services

to be achieved and a minimum of customer-relevant

residual errors to be present at the end of the project.

All projects for which either the scope of services has

been reduced, the project duration has been extended

or an unreasonably high number of residual errors are

present at the end of the project are therefore

classified as unsuccessful.

4.2 Analysis of Critical Success Factors

In addition to the examination of the project's success,

we analysed the relevant literature between 1960s and

today with regard to the concept of critical success

factors for projects. Again, we clustered and

examined the findings from section 2 according to

their relative frequency (figure 3). Results show, that

the existence of the right competences and skills, a

ICAART 2021 - 13th International Conference on Agents and Artiﬁcial Intelligence

526

clear goal and realistic planning are fundamental

factors for successful project implementation. In

addition, more than half of the researchers surveyed

consider top management support and effective

communication to be decisive success factors. Other

relevant characteristics were an understanding of

customer needs, the availability of resources and the

existence of common management systems (project

controlling, change and risk management system).

Interestingly, if the range of the analysis is reduced to

the period of 2010 and after, the set of success factors

is almost identical. While top management support is

no longer one of the ten most cited success factors,

social and political factors are among the second most

cited criteria. Realistic project planning is still seen as

a relevant instrument for project implementation.

More than 90% now also name the implementation of

a risk management system as a critical success factor.

Figure 3: Top cited CSFs ranked by relative frequency.

A possible explanation for this could be the

increasing and rapidly changing demands of

customers, competition and politics. This can also be

measured by the fact that effective project controlling

and change management are becoming increasingly

important. Good communication and an

understanding of customer needs are now enumerated

by 75% of the researchers. After all, almost six out of

ten researchers consider clear project goals and the

availability of resources such as budget, technology

and logistics to be decisive success factors.

In conclusion, this analysis shows that many of the

identified CSFs seem to have cross-project

applicability. One possible explanation for this is that

CSFs embody a success-oriented definition of the

project framework or project environment and are

only indirectly linked to the operative results of the

project. In addition, a large number of possible

influencing factors were found that could serve as

possible input features.

4.3 Evaluation of the Employee Survey

In addition to the broad literature research, we

conducted an employee survey in the research area.

Results show that the project success indicators found

in our literature research were confirmed within the

scope of this survey. Accordingly, 86% of the

participants agreed with the assertion that the

technical specifications of a vehicle development

project can have an influence on the success of the

project’s progress. Three-quarters of the respondents

confirmed efficiency in implementation as a success

indicator, and just over half attribute a major role to

product quality and employee satisfaction. In

addition, 43% of those surveyed rated the scope of

services as an indicator of project success. Only 14%

of the respondents rated the planned duration of the

projects as the least strongly involved in the success

of the project. The reason for this lies in the planned

target development time of a vehicle project that

always has a constant length of 48 months.

Following the introductory questions on the success

of the project, the participants were confronted with

the objectives and deadlines. It can be seen from the

survey’s results that a strict and detailed project

schedule is kept up to date in less sub-projects.

Furthermore, the assertion from previous expert

interviews could be proven quantitatively (95%) that

delays in the awarding of suppliers have a negative

effect on a timely development. Also, 95% of the

participants agreed that on the one hand the risk of

later milestone shifts increases considerably if

deadlines are not met in the early development phase.

On the other hand, late concept decisions have a

negative impact on the deadline targets of vehicle

development projects as well. More than three-

quarters of the respondents also stated that process

adherence in the early project phases can be a success

factor with regard to deadline targets.

Furthermore, we asked for specific critical project

success factors in the closer meaning. Based on the

responses of the participants (62%), the research area

is currently less able to learn from past mistakes. In

contrast, there was predominantly indifference as to

whether project risks in the research area were

identified early, evaluated and documented with

measures and pursued. Feedback on the question of

top management support for acute challenges

continued to be widely distributed. 69% of

respondents believe that the use of proven

technologies would facilitate the development of

vehicle projects. This can be an indication that

different vehicle development projects are also based

on different factors influencing project success.

Predicting the Progress of Vehicle Development Projects: An Approach for the Identiﬁcation of Input Features

527

However, most of the CSF already identified in

theory could also be quantitatively confirmed by the

study. As a result, 69% of respondents estimate that

the use of proven technologies would facilitate

vehicle development. 90% of the participants agreed

that good performance in cooperation with suppliers

contributes to the success of the project. Furthermore,

81% of the participants agreed that the number of

functions have an influence on product quality, e.g.

the number of customer-relevant errors. This is in

contrast to the few mentions that the scope of services

can influence the success of the project. Absolute

agreement (100%) prevails that both the mere number

of new functions and new technologies in vehicle

architecture or new highly networked components

always present a major challenge for the planned

development and validation. Conversely, this also

means that the probability of success is higher for

projects in which a large number of components

and/or functions are taken over from other projects.

With 90% of votes in each case, the general employee

satisfaction and a stable, departmental internal policy

can be counted as CSF. The size and complexity of

the project was also clearly mentioned (100%) as a

CSF.

Table 1: Risk factors for the project success.

Finally, the interviewees had the optional opportunity

to name further factors, which from their individual

point of view could risk the timely development and

safeguarding of vehicle projects. The identified risks

were gathered in table 1 and assigned to the defined

project success indicators. Interestingly, late

decisions, either with a general reference to the

project or already when awarding contracts to

suppliers, appear to represent a major risk for the

schedule targets. Additionally, the progress of the

function build-up or the functional perceptibility can

be an important indicator of project’s progression.

Furthermore, we learnt from the survey that

downgrading projects in the decision-making

processes influence the quality of the products

negatively. Also, we identified the change of

suppliers in the further development of ECUs as a

relevant risk. A surplus of reporting and task force

operations appears to have a negative effect on

employee satisfaction. Finally, with unclear goals,

inadequate communication, missing cooperation

models, too few resources and high bureaucratic

hurdles, many CSFs known from theory were once

again recognized.

Principally, the identified factors show a good

applicability for machine learning models. Among

the examples mentioned above, we can identify

numerical data, meaning continuous or discrete data,

e.g. the average number of persons per project,

number of shifts in the perceptibility of functions and

so on. Furthermore, there is a lot of categorical data,

e.g. exchange of suppliers for the advancement of an

ECU, project size and so on. And finally, we can

make use of a lot of time series data. After initial

analysis of the data, it can be confirmed that this

information is available in a sufficiently large

quantity and is complete except for a few gaps. Using

feature engineering methods, we will address this

problem later. This means that the data will provide

the basis for e.g. approaches of multivariate

prediction problems. Furthermore, we are able to use

the labelled information for classification problems.

Depending on the specific issue, the data can also be

used for clustering purposes. From this it can be

concluded that the information available to us will

provide a very high usability for the application in

machine learning approaches.

5 CONCLUSIONS

With this paper, we have presented a procedure to

identify measurable factors that influence the course

of vehicle development projects in order to use them

as input features for machine learning models to

predict project’s progression. To achieve this goal, a

broad literature research about applied machine

learning was conducted. It turned out that there is still

a large research gap on machine learning applications

in operative project management. Furthermore, the

relevant project management literature was analysed

with regard to the different models of project success

and critical success factors in projects. As a result of

these investigations, the most frequently cited project

success indicators and CSFs have been identified.

The overviews of the indicators of project success and

the CSFs that emerged from the literature research do

not only provide a valuable list of possible

influencing factors. They also provide valuable

information that are extremely helpful to project

ICAART 2021 - 13th International Conference on Agents and Artiﬁcial Intelligence

528

managers from all industry sectors in the initialization

and definition phase of a variety of projects.

At the same time, an analysis of the project success

and the CSFs provided a more precise view of the use

case. The results of the literature research could be

confirmed for the use case. These results were then

used to develop a working definition for the success

of the project in the implementation phase. With

regard to the applicability as input features in

machine learning models these findings were

consolidated. Finally, concrete examples from the use

case were assigned to the identified success indicators

(figure 4). In addition to the identified CSFs, a large

set of influencing factors on the course of vehicle

development project in the research area has been

identified.

Figure 4: Extract of influencing factors on vehicle

development project’s progressions.

Our long-term objective is to develop a general

framework for the prediction of development

project’s progressions. The basic idea of the planned

procedure is shown in figure 5. Hence, the next step

is to build an adequate DataSet containing the

identified input features. Therefore, it must be

checked whether the data are quantitatively recorded

and whether accessing can be established.

Furthermore, it has to be clarified how to deal with

information that is not yet recorded. Here we expect

to have data from different databases and about

various car projects. We will have access on general

project-relevant information (e.g. start and end dates,

milestones, number of functions, size and complexity,

amount of ECUs, technical specifications, soft facts,

etc.), time series data about function releases and

error handlings (e.g. car project, ECU, responsible

department, time to build-up functions, time to bug

fix, market, etc.) and additional data (e.g. date of

contracts with suppliers, employee satisfaction, etc.).

As projects are characterized by their uniqueness, it is

assumed that the effects of the input features on the

course of the project can vary between the various

projects. Based on this, the influencing factors in

relation to the vehicle projects must be examined

precisely. The findings from section 3 support this

assumption, since we found out, that there may be

different factors that can possibly influence a vehicle

development project. Therefore, we will analyse

whether the influence of the features is different or

not at different phases in the course of the project. For

this reason, we will apply feature selection methods

on our data (with focus on filter and wrapper

methods) and compare the results with those from this

paper to generate a dataset for further use.

Figure 5: ML-Approach for predicting project maturity.

Within the research area, automotive

development projects are mainly divided into two

categories: SOP and LCM projects. Based on domain

knowhow there is evidence that in many cases the

course of SOP and LCM projects seems to be

different. This implies different factors that influence

the course of the project. To predict the progression

of a project, it is therefore valuable to know

beforehand whether the project's progression to be

predicted is an SOP or LCM project. In this context,

an approach for the classification of vehicle

development projects shall be developed, taking into

account common methods (e.g. kNN, LSTM,

Random Forest, SVM, ...).

Finally, we plan on implementing different models

for time series forecasting (e.g. random forests,

recurrent neural networks, convolutional neural

networks, …), in order to predict the further course of

the project. Special attention shall be given to the

prediction of the function build-up and the error

reduction. The model that represents the best

performance based on the selected evaluation metrics

will then be selected. For a milestone to be defined, a

measurement is then to be made as to whether the

vehicle project can withstand the requirements of this

milestone and whether it can be classified as

successful. Finally, the model will be verified on the

use case of a present vehicle development project

with regard to its effectiveness by making predictions

about the project’s progressions.

Predicting the Progress of Vehicle Development Projects: An Approach for the Identiﬁcation of Input Features

529

REFERENCES

Ahmed, A. A., Musonda, I., Pretorius, J. H. and Hamasha,

M., 2019. Natural logarithm transformation for

predicting procurement time of PPP projects in Nigeria.

In: Cogent Engineering (6(1))

Atkinson R., 1999. Project management. Cost, Time and

Quality, two best Guesses and a Phenomenon, it’s Time

to Accept other Success Criteria. In: International

Journal of Project Management (17(6)), pp. 337–342

Aytug, H., Bhattacharyya, S., Koehler, G. J. and Snowdon,

J. L., 1994. A review of machine learning in scheduling.

In: IEEE Transactions on Engineering Management

(41(2)), pp. 165–171

Baccarini, D., 1999. The Logical Framework Method for

Defining Project Success. In: Project Management

Journal (30(4)), pp. 25–32

Belassi, W. and Tukel, O. I., 1996. A New Framework for

Determining Critical Success/Failure Factors in

Projects. In: International Journal of Project

Management (14(3)), pp. 141–151

Boetticher, G. D., 2001. Using Machine Learning to Predict

Project Effort. Empirical Case Studies in Data-Starved

Domains. In: First international workshop on model-

based requirements engineering

Boynton, A. C. and Zmud, R. W., 1984. An Assessment of

Critical Success Factors. In: Sloan Management Review

(25(4)), pp. 17–27

Deloitte, 2019. Global Automotive Consumer Study 2019 -

Advanced vehicle technologies and multimodal

transportation - Global Focus Countries, retrieved May

01, 2019 from:

https://www2.deloitte.com/content/dam/Deloitte/de/D

ocuments/consumer-industrial-products/Deloitte-

global-automotive-consumer-study-2019-focus-

countries.pdf

Fortune, J. and White, D., 2006. Framing of project critical

success factors by a systems model. In: International

Journal of Project Management (26), pp. 53–65

Hassan, H., Aue, A., Chen, C., Chowdhary, V., Clark, J. et

al., 2018. Achieving Human Parity on Automatic

Chinese to English News Translation, retrieved October

01, 2019 from:

https://www.microsoft.com/en-

us/research/uploads/prod/2018/03/final-achieving-

human.pdf

Hu, Y., Huang, J., Chen, J., Liu, M. and Xie, K., 2007.

Software Project Risk Management Modelling with

Neural Network and Support Vector Machine

Approaches. In: Third International Conference on

Natural Computation (ICNC 2007)

Huang, E. and Chen, S.-J. (Gary), 2006. Estimation of

Project Completion Time and Factors Analysis for

Concurrent Engineering Project Management. A

Simulation Approach. In: Concurrent Engineering

(14(4)), pp. 329–341

Jugdev, K. and Müller, R., 2005. A Retrospective at Our

Evolving Understanding of Project Success. In: Project

Management Journal

Kermany, D. S., Goldbaum, M., Cai, W., Valentim, C. C.

S., Liang, H. et al., 2018. Identifying Medical

Diagnoses and Treatable Diseases by Image-Based

Deep Learning. In: Cell (172(5)), pp. 1122-1131

Kometa S., Olomolaiye P. O. and Harris F. C., 1995. An

Evaluation of Clients’ Needs and Responsibilities in the

Construction Process. In:

Engineering, Construction

and Architectural Management, pp. 45–56

Kumaraswamy, M. M. and Thorpe, A., 1996. Systematizing

Construction Project Evaluations. In: Journal of

Management in Engineering (12(1)), pp. 34–39

Li, J., Sun, M., Han, D., Wang, J., Mao, X. and Wu, X.,

2019. A knowledge discovery and reuse method for

time estimation in ship block manufacturing planning

using DEA. In: Advanced Engineering Informatics

(39), pp. 25–40

Müller, R. and Jugdev, K., 2012. Critical success factors in

projects. Pinto, Slevin, and Prescott – the elucidation of

project success. In: International Journal of Managing

Projects in Business (5(4)), pp. 757–775

Navarre, C. and Schaan, J. L., 1990. Design of Project

Management Systems from Top Management’s

Perspective. In: Project Management Journal

(Vol.XXI(2)), pp. 19–27

Pedroso, M.: Application of Machine Learning Techniques

in Project Management Tools, 2017

Perini, A., Susi, A. and Avesani, P., 2013. A Machine

Learning Approach to Software Requirements

Prioritization. In: IEEE Transactions on Software

Engineering (39(4)), pp. 445–461

Pinto, J. K. and Slevin, D. P., 1986. The Project

Implementation Profile. New Tool for Project

Managers. In: Project Management Journal (17(4)), pp.

57–70

Pinto, J. K. and Slevin, D. P., 1988. Project Success:

Definitions and Measurement Techniques. In: Project

Management Journal (19(1)), pp. 67–72

Pinto, M. B. and Pinto, J. K., 1991. Determinants of Cross-

functional Cooperation in the Project Implementation

Process. In: Project Management Journal

(Vol.XXII(No.2)), pp. 13–20

Sharma, M., Bedi, P., Chaturvedi, K. K. and Singh, V. B.,

2012. Predicting the priority of a reported bug using

machine learning techniques and cross project

validation. In: 12th International Conference on

Intelligent Systems Design and Applications (ISDA)

Shenhar, A. J., Levy, O. and Dvir D., 1997. Mapping the

Dimensions of Project Success. In: Project

Management Journal (June), pp. 5–13

Srinivasan, K. and Fisher, D., 1995. Machine learning

approaches to estimating software development effort.

In: IEEE Transactions on Software Engineering

(21(2)), pp. 126–137

Wuellner, W. W., 1990. Project Performance Evaluation

Checklist for Consulting Engineers. In: Journal of

Management in Engineering (6(3)), pp. 270–281

ICAART 2021 - 13th International Conference on Agents and Artiﬁcial Intelligence

530