Syddansk Universitet Neural Networks to model the innovativeness perception of co-creative firms

Sana	23.09.2017
Hajmi	194.4 Kb.
	#16332

General rights
Take down policy

Syddansk Universitet

Neural Networks to model the innovativeness perception of co-creative firms

Tanev, Stoyan

Published in:

Expert Systems with Applications

DOI:

10.1016/j.eswa.2012.05.022

Publication date:

2012

Document version

Early version, also known as pre-print

Citation for pulished version (APA):

Tanev, S. (2012). Neural Networks to model the innovativeness perception of co-creative firms. Expert Systems

with Applications, 39(16), 12719-12726. DOI: 10.1016/j.eswa.2012.05.022

General rights

Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners

and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research.

• You may not further distribute the material or use it for any profit-making activity or commercial gain

• You may freely distribute the URL identifying the publication in the public portal ?

Take down policy

If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately

and investigate your claim.

Download date: 23. sep.. 2017

Review

Neural Networks to model the innovativeness perception of co-creative ﬁrms

Giacomo di Tollo

⇑

, Stoyan Tanev

, De March Davide

, Zheng Ma

LERIA, University of Angers, 2, Boulevard Lavoisier, 49045 Angers Cedex 01, France

Integrative Innovation Management Unit, Institute of Technology and Innovation, University of Southern Denmark, Niels Bohrs Allé 1, DK-5230 Odense M, Denmark

EvoSolutions S.r.l., Viale Ancona 17, 30172 Venezia, Italy

Department of Molecular Sciences and Nanosystems, Ca’ Foscari University of Venice, Italy

Integrative Innovation Management Unit, Institute of Marketing Management, University of Southern Denmark, Niels Bohrs Allé 1, DK-5230 Odense M, Denmark

a r t i c l e

i n f o

Keywords:

Innovation complexity

Value co-creation

Self-Organising Maps (SOM)

Artiﬁcial Neural Networks (ANN)

a b s t r a c t

Value co-creation is an emerging business, marketing and innovation paradigm describing the ﬁrms apti-

tude to adopt practices enabling their customers to become active participants in the design and devel-

opment of personalised products, services and experiences. The main objective of our contribution is to

make a quantitative analysis in order to assess the relationship between value co-creation and innovation

in technology-driven ﬁrms: we are using Artiﬁcial Neural Network (ANN) to investigate the relationship

between value co-creation and innovativeness, and Self Organising Map (SOM) models to cluster ﬁrms in

terms of their degree of involvement in co-creation and innovativeness. Results from the ANN show that a

strong relationship exists between value co-creation and innovativeness; furthermore, SOM are well per-

forming in identifying cluster of ﬁrms that are more involved in co-creation values. Our work makes a

methodological contribution by adopting and validating a combination of techniques that is able to

address complexity and emergence in value co-creation systems.

1. Introduction

The ongoing globalisation processes and the emergence of mar-

kets with heterogeneous customer needs are forcing companies to

look for new sources of innovation and competitive differentiation,

e.g., by shifting their classical frame of mind to end users through

the adoption of a new marketing approach in which end users be-

come an active part in designing and shaping personalised prod-

ucts, services and experiences. This trend has established value

co-creation (

Chesbrough, 2011; Mindgley, 2009b; Ramaswamy &

Gouillart, 2010b; Tanev, Knudsen, & Gerstlberger, 2009

) as an

important marketing and innovation paradigm (

Lusch & Vargo,

2006; Prahalad & Krishnan, 2008; Prahalad & Ramaswamy,

2004

), describing how customers and end users could be involved

as active participants in the process of value creation (

Etgar, 2008;

Prahalad & Ramaswamy, 2004; Payne, Storbacka, & Frow, 2008

The concept of co-creation is based on the design and develop-

ment of customer participation platforms, which allow ﬁrms to use

technological and human resources to beneﬁt from the engage-

ment of individuals and communities (

Nambisan & Baron, 2009;

Nambisan & Nambisan, 2008; Sawhney, Gianmario, & Prandelli,

2005

). These platforms enable the personalisation of new products

and services challenging traditional marketing segmentation tech-

niques, by promoting a new service-dominant logic (

Vargo &

Lusch, 2004; von Hippel, 2006a

) which allows ﬁrms to tackle het-

erogeneous markets and to better ﬁt the customer’s needs.

The innovation-related implications of value co-creation are

emerging amongst the most relevant topics in value co-creation re-

search (

Bowonder, Dambal, Kumar, & Shirodkar, 2010; Kristenson,

Matthing, & Johansson, 2008; Michel, Brown, & Gallan, 2008;

Midgley, 2009a; Nambisan & Baron, 2009; Prahalad & Ramaswam-

y, 2003; Prahalad & Krishnan, 2008; Roberts, Bake, & Walker, 2005;

Romero & Molina, 2009; Sawhney et al., 2005; Tanev et al., 2009

)

since the new paradigm entails a new vision about the relationship

between marketing and innovation. Nevertheless, most of the

works have been focused on qualitative case studies, emphasising

the role of the customer participation in co-creation on the innova-

tion outcomes, such as innovation cost, time-to-market, new prod-

uct or service quality and development capacity (

Bowonder et al.,

2010; Kristenson et al., 2008; Midgley, 2009a; Nambisan & Baron,

2009; Nambisan, 2009; Prahalad & Krishnan, 2008; Ramaswamy &

Gouillart, 2010a; Romero & Molina, 2009

). It has to be remarked

that the performance of co-creation practices is measured from

an innovation perspective alone, neglecting side effects such as

brand perception, customer satisfaction, or customer-ﬁrm rela-

tionship quality (

Nambisan & Baron, 2009; Nambisan, 2009

). In a

nutshell, existing literature fails in analysing the emerging nature

of value co-creation systems, neglecting complexity and emer-

gence would affect business model design, pricing models and

management practices (

Desai, 2010; Ng, 2010; Tanev et al.,

2009

). There are no sound quantitative studies focusing on the

http://dx.doi.org/10.1016/j.eswa.2012.05.022

⇑

Corresponding author.

E-mail address:

giacomodt@gmail.com

(G. di Tollo).

Expert Systems with Applications 39 (2012) 12719–12726

Contents lists available at

SciVerse ScienceDirect

Expert Systems with Applications

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / e s w a

co-creative sources of innovation. A ﬁrst investigation of this spe-

ciﬁc aspect was performed by

Tanev et al. (2011)

, focusing on on-

line Internet data and linear regression analysis to examine the

relationship between the degree of involvement of ﬁrms in value

co-creation activities and the frequency of their online comments

about their new products, processes and services. However, the ap-

proach suggested therein has an obvious limitation: it relies on a

linear relationship between co-creation and innovation, which

cannot be assumed without loss of generalisation. It is worthwhile

exploring more general models that could potentially take into ac-

count the complexity and emergence in value co-creation systems.

The purpose of our work is to perform a quantitative analysis by

enhancing the approach by

Tanev et al. (2011)

w.r.t. the aforecited

limitation. We will use Artiﬁcial Neural Network (ANN) models to

examine the relationship between the degree of ﬁrms’ value co-

creation activities and the frequency of online comments about

new products, processes and services. The relevance of adopting

the ANN approach stays in that they do not make any assumption

about the relationship under study. By performing this analysis we

are aimed to test the hypothesis that: ﬁrms with a higher degree of

involvement in co-creation activities have a better opportunity to

articulate the innovative features of their new products, processes

and services. Furthermore, we will classify ﬁrms w.r.t. the degree

of their co-creation activities by means of Self-Organising Map

(SOM). This operation will enable us to identify the ﬁrms that are

most active in co-creation, and will open up the possibility for fu-

ture qualitative research focusing on the distinctive features of dif-

ferent co-creation components as part of emerging co-creation

strategies.

Testing the above hypothesis will provide insights to under-

stand the context of an increasingly globalised competitive envi-

ronment, in which ﬁrms are facing the limits of traditional

marketing techniques that do not necessarily lead to a better com-

petitive positioning or differentiation (

Prahalad & Krishnan, 2008;

Prahalad & Ramaswamy, 2004

The two key contributions of this paper can be therefore sum-

marised as follows:

applying ANNs to model the relationship between value co-cre-

ation and innovation;

using SOM to classify ﬁrms in terms of the degree of their

involvement in co-creation and innovation.

ANNs and SOMs are considered respectively as supervised

(

Reed & Marks, 1999

) and unsupervised (

Hinton & Sejnowski,

1999

) Neural Network approaches, able to adapt their topologies

and parameters in order to minimise some pre-deﬁned measures

of goodness (usually root mean squared error in ANN approach

and Euclidean distance in SOM). Up to the authors knowledge, this

is the ﬁrst application of these approaches to innovation and co-

creation. This paper is organised as follows. Section

provides a

conceptual discussion of value co-creation within the context of

innovation and complexity theory, Section

describes the statisti-

cal models applied in this paper, which will be used to explain the

correlation between co-creation and innovation. Section

provides

results describing the relationship between value co-creation and

innovation including a comparison with previous works. Final re-

marks are presented in Section

2. Value co-creation, innovation, and complexity

The adoption of value co-creation practices challenges the tradi-

tional ways of innovation management by promoting a new vision

of innovation itself (

Kristenson et al., 2008; Prahalad & Krishnan,

2008; Tanev et al., 2009

). The new co-creative vision of innovation

relies on two key distinctive features. The ﬁrst one is the customer-

driven aspect of the value co-creation activities, in which value co-

creation platforms can be seen as a natural extension of some key

aspects of user-driven innovation initiatives (

von Hippel, 2006a,

2006b

) by focusing on the development of participation platforms

(

Nambisan & Nambisan, 2008; Nambisan & Baron, 2009; von Hip-

pel, 2001

) and by searching for lead users (

Droge, Stanko, & Pollitte,

2010; von Hippel, 2006b

); the latter distinctive feature is the focus

on a balance between cooperation and competition, or co-opetition.

The co-opetitive dimension of value co-creation platforms leads

to deﬁne a more dynamic scenario of the economic mechanisms

which trigger the innovation processes. These mechanisms operate

on the basis of multiple transactions between customers, partners

and suppliers, at multiple access points across the value network.

They enable customers and end users to control the relationship

between price and user experience (

Etgar, 2006; Prahalad &

Ramaswamy, 2004

) by providing them the opportunity to create

speciﬁc value chain conﬁgurations leading to new value compo-

nents, new ways of using existing solutions, or the radical

improvement of an existing product or service (

Bowonder et al.,

2010; Kristenson et al., 2008

). In this context customers are re-

ferred to as innovators and co-creators.

The participatory platform nature of value co-creation practices

enables a broader and more systematic positioning of customers

and end users across the entire innovation lifecycle leading to a

signiﬁcant enhancement of the user-driven innovation potential.

As a result, the development of value co-creation platforms is

increasingly recognised as a promising innovation strategy

(

Bowonder et al., 2010; Nambisan & Baron, 2009; Nambisan,

2009; Midgley, 2009a; Prahalad & Ramaswamy, 2003; Romero &

Molina, 2009

The co-creation paradigm associates the source of value with

the co-creation experience which is actualised through the com-

pany-customer interaction events. By co-creating with the net-

work, the customer becomes an active stakeholder who can

deﬁne the type of interaction and the speciﬁc personal context of

the encountering event (

Prahalad & Ramaswamy, 2003

). The per-

sonal nature of the interaction enables the emergence of new value

dimensions which are based on the quality and the personal rele-

vance of the interaction events as well as on the opportunity for

customers to co-create unique end products, services and experi-

ences. These new dimensions are important for the emergence of

experience innovation networks putting the individual at the heart

of co-creation experience through the dynamic shaping of techno-

logical-business process and human resource infrastructures

(

Prahalad & Krishnan, 2008

). In this sense, the value co-creation

paradigm represents a speciﬁc market-driven approach to the

adoption of an open innovation business philosophy.

Eventually, the adoption of value co-creation practices could

pave the way for the emergence of disruptive innovation business

models (

Christensen, 2006

). Some of the sources for such opportu-

nities are: technological breakthroughs as enablers of efﬁcient co-

creation mechanism; changes in the industry logic leading to the

emergence of new channels for reaching customers; changes in

customer preferences and lifestyles (

Payne et al., 2008

). The dis-

ruptive innovation potential of value co-creation-driven business

models represents a great opportunity for future research.

The adoption of a value co-creation business philosophy re-

quires a re-conceptualisation of the common sequential under-

standing of the value chain into a complex and dynamic network

of value, producing relations between producers, suppliers, cus-

tomers and end users. Some scholars still use the usual linear ap-

proach to business driven by a Newtonian or mechanistic view of

the world, in which each product is associated to a value, is pro-

duced away from the market, and is sold by means of decisions

made to maximise the proﬁt. Customers are cut out of any

12720

G. di Tollo et al. / Expert Systems with Applications 39 (2012) 12719–12726

opportunities to become part of the innovation process. In contrast,

one could use a quantum physics view to describe how companies

move away from a company-and-product centric approach to par-

ticipation platform oriented approaches where customers, partners

and suppliers cooperate and co-create value (

Tanev et al., 2009

This shift implies a new value chain model which is open, non-lin-

ear, operationally parallel, and three dimensional. The features of

such a quantum physics model can be summarised as follows: ﬁrst,

there is an uncertainty principle since the speciﬁc value chain tra-

jectory is not known in advance; second, there is a complementarity

principle since the output of a particular value chain conﬁguration

could be considered in two different and complementary ways as

an end product or solution, or as a platform with a focus on the

critical role of its network or partnership enabled value compo-

nent; third, the power of a value co-creation platform is deter-

mined by summing all potential multiple path conﬁgurations of

value when calculating the probability for a speciﬁc market offer;

fourth, the role of the observer (the customer) is critical to the spe-

ciﬁc nature of the ﬁnal outcome of the interaction; ﬁfth, there is a

place for the manifestation of non-local phenomena such as net-

work knowledge and collective wisdom that provides an additional

value dimension making a value co-creation platform more com-

petitive (

Tanev et al., 2009

In this model the current pricing mechanisms may not be effec-

tive and should involve the development of more dynamic system

level pricing, a better understanding of system capacity, with the

system as a unit of analysis in prescribing innovatively different

pricing schemata (

Ng, 2010

). Systems thinking, complexity theory,

network and system sciences will deﬁnitely impact the pricing

models of value co-creation systems because of the multiple-

agent-based interconnected nature of the market offerings. The

nature of the interdependencies is accelerated by technologies

moving towards convergence, resulting in the involvement of mul-

tiple stakeholders and multiple customers who are all contributing

resources into the system while, at the same time, paying for dif-

ferent parts and deriving different beneﬁts from it (

Ng, 2010

Adaptive leadership and management practices will be crucial

for the emergence of competitive value co-creation networks.

Organizations interested in adopting value co-creation strategies

need people who could master the principles of adaptive leader-

ship within customers, suppliers and internal networks (

Desai,

2010; Prahalad & Ramaswamy, 2004

2.1. Research methodology

The development of business insights by using unstructured

public data becomes increasingly popular resource for both schol-

ars and practitioners.

Hicks, Libaers, Porter, and Schoeneck (2006)

and

Ferrier (2001)

pioneered the concept that an analysis of the

frequency of speciﬁc keywords on public websites and corporate

news releases can be an adequate representation of the degree of

importance placed by ﬁrms on the activities represented by those

keywords.

More recently

Allen, Tanev, and Bailetti (2009)

demonstrated

that such methodology could be used to classify value co-creation

practices and formalised the key steps of the data gathering and

analysis work ﬂow, showing that the frequencies of a speciﬁc set

of keywords can be used to extract the key components of value

co-creation activities, using those ideas to outline a detailed re-

search process. This process starts with a careful construction of

a set of keywords to represent the different value co-creation con-

stitutive dimensions. Then, the frequency of use of each of the key-

words on companies’ websites and news releases is measured. This

procedure is justiﬁed by the fact that most co-creation activities

undertaken by technology-driven ﬁrms are performed or described

online, since the more a ﬁrm describes a speciﬁc activity, the more

it deems this activity relevant for its current situation (

Allen et al.,

2009; Ferrier, 2001; Hicks et al., 2006

). Principal Component Anal-

ysis (PCA) is then used to identify emerging groups of keywords

that could be associated with speciﬁc self-consisting groups of

activities (components). Last, a heuristic technique is outlined to

classify ﬁrms with regards to their involvement in the different

co-creation activities (

Allen et al., 2009

) by ranking ﬁrms with re-

gards to each of the co-creation components.

This methodology was further enhanced by

Tanev et al. (2011)

examining the perception of ﬁrms’ innovativeness by measur-

ing the frequency of ﬁrms’ online comments about their new

products processes and services;

applying linear regression analysis to test the existence of a

positive association between the degree of ﬁrms’ involvement

in value co-creation and the perception of their innovativeness;

using k-means cluster analysis to classify the ﬁrms in terms of

the degree of their involvement in value co-creation.

In our work, we are using the same research sample, the same

co-creation components and the same innovation metric by

Tanev

et al. (2011)

, though we want to enhance its research, applying non

linear models and adaptive paradigms, that could be better suited

for the new dynamics of the co-creation philosophy, by:

applying ANN approach to model the relationship between co-

creation and innovation;

using the SOM technique to classify the ﬁrms in terms of the

degree of their involvement in co-creation and innovation.

The composition of the co-creation components derived by

Ta-

nev et al. (2011)

is based on dataset including 273 ﬁrms selected

amongst cases found in the reviewed value co-creation literature

and amongst ﬁrms engaged in OSS projects. Firms have been se-

lected amongst members of the Eclipse OS Foundation and from

two websites: Open Source Experts –

www.opensourceex-

perts.com

, and the Canadian Companies Capabilities Directory of

OS Companies –

http://strategis.ic.gc.ca/epic/site/ict-tic.nsf/en/

h_it07356e.html

). A summary of the dataset features can be found

Table 1

. In Sections

2.2 and 2.3

we are going do describe the

matrices used to model co-creation and innovation. These metrics

will be used for the experimental analysis in what follows.

2.2. Value of co-creation components

As stated in Section

2.1

, we need to deﬁne co-creation compo-

nents in order to test if there is a relationship between these com-

ponents and the innovation metric. We have chosen to use three

components deﬁned in

Tanev et al. (2011)

Based on these results, The ﬁrst co-creation component is re-

ferred to as Resources and processes and is interpreted as (resources,

processes, tools and mechanisms) enabling (customer and user

involvement)

(production,

assembly,

manufacturing

and

self-service) aiming at (design and process ﬂexibility) based on

Table 1

Firms included in the research sample: the label GEN indicates general type (non-

software) ﬁrms, ECL indicates ﬁrm from the Eclipse Foundation, OSS indicates open

source software ﬁrms not related to the Eclipse Foundation. Column Freq indicates the

cardinality of each type set, and Percent their relative percentage.

Type of ﬁrms

Frequency

Percent

GEN

23.8

ECL

133

48.7

OSS

27.5

Total

273

100.0

G. di Tollo et al. / Expert Systems with Applications 39 (2012) 12719–12726

12721

(product modularity and sharing of internal expertise, resources

and IP).

The second co-creation component is referred to as Customer

relationships and is interpreted as (customer relationships) enabled

through (partnerships and cooperation) aiming at (cost reduction,

design and process ﬂexibility), and leading to (better customer

and end user experiences) based on (risk management, transpar-

ency and trust).

The third co-creation component is referred to as Mutual learn-

ing and is interpreted as (mutual learning mechanisms) based on

the existence of user networking forums enabling (customer sugges-

tions, input, demands and requests), and leading to (multiple op-

tions for users) through involvement in (test and beta trials).

Table 2

shows the main descriptive statistics of the three co-crea-

tion variables that were constructed by adding up the ratings of

each keyword weighted by its loading (see

Tanev et al. (2011)

for

further details about this procedure).

2.3. Online innovation metric

Table 3

shows the descriptive statistics of the normalised inno-

vation metric assessing the ﬁrm’s own Perception of innovative-

ness. It was used by

Tanev et al. (2011)

by measuring the

frequency of ﬁrms’ online comments about new products, services

and processes and it is composed in a way that it would detect any

online statement containing the combination of the words new and

product, or the words new and service, or new and process etc. (the

complete boolean expression is: new ^( product _ service _ pro-

cess _ application _ solution _ feature _ release _ version _ launch

_ introduction _ introduce _ ‘‘new product’’ _ ‘‘new service’’ _

‘‘new process’’).

It should be pointed out that the online frequency of the inno-

vation keyword is not a traditional innovation metric since it does

not account directly for the number of new products, processes and

services but rather the frequency of online comments about their

new features. This new metric embeds the advantage of emphasis-

ing the ability of a ﬁrm to differentiate itself by articulating the

innovative aspects of its products and services. The introduction

of such metric could help ﬁrms in addressing the old paradigm of

an increasing product variety coexisting with a decreasing cus-

tomer satisfaction (

Prahalad & Ramaswamy, 2004

3. Statistical approaches for a quantitive analysis of innovation

and co-creation

This section presents respectively a supervised approach to ﬁt

and evaluate the relation between our innovation metrics and va-

lue co-creation (Section

3.1

) and an unsupervised approach to clus-

ter co-creative ﬁrms in terms of their involvement in co-creation

values (Section

3.2

). We extend the previous work by

Tanev et al.

(2011)

by applying a Neural Network model to assess the relation-

ship between the innovation and co-creation variables and a self

organizing map to classify the ﬁrms of our case study.

3.1. A Neural Network approach to model innovation based outcomes

Artiﬁcial Neural Networks (ANN) are computing methods (algo-

rithms) whose behaviour mimics the human brain (

Hykin, 1999;

Angelini, di Tollo, & Roli, 2008

). ANNs are composed of basic ele-

mentary units (neurons) which, when taken as single units, are

able to execute some simple basic operations, but when connected

to create a network, they can perform complicated tasks and solve

complex problems, especially when the particular problem model

is unknown in advance and when the relationships amongst the

different components are non-linear. The most common ANN mod-

el is the multi-layer Neural Network, often called feed-forward

Neural Network, which allows a generalisation of the model

becoming a general function approximator. The feed-forward Neu-

ral Network introduced some layers between the input variables X

and the observed response Y. These layers are called hidden layers

and each of that can assume different number of neurons and dif-

ferent activation functions. Each neuron is connected to each neu-

ron belonging to an adjacent layer, while there are no connections

between neurons of the same layer. The connections are called

weights and represent the parameters of the Neural Network.

These parameters can change their values in a learning procedure

and this adaptation of weights allow the model to better ﬁt with

the observed output. We refer to

Bishop (1996)

for a wider expla-

nation of the Neural Network models and to

Rumelhart, Hinton,

and Williams (1986)

for the description of the most applied opti-

misation algorithm, the Back-Error Propagation, for estimating

the weights. The one hidden layer architecture is wildly used be-

cause it can approximate any function with a ﬁnite number of dis-

continuities, arbitrarily well, given sufﬁcient neurons in the hidden

layer (

Hagan, Demuth, & Beale, 1996

The main advantage of ANN approaches consists in their gener-

alisation capabilities, i.e. in their ability to operate over data that

have never been seen before, and for this reason they are used in

tasks such as pattern recognition, forecasting, optimisation and

classiﬁcation (

Angelini et al., 2008; Zemella, De March, Borrotti,

& Poli, 2011

). In addition, the application of the ANN approach

has another signiﬁcant advantage in not relying on any speciﬁc

preliminary model. Furthermore, they are robust with respect to

noisy and missing data, which do not hinder the network opera-

tions (but of course trigger some degree of tolerable performance

degradation). All those requirements make their use appropriate

for the problem at hand, in which a model is still far from being

developed.

In this research we consider the three value co-creation compo-

nents, deﬁned in Section

2.2

, as the input variables and the percep-

tion of innovation, deﬁned in Section

2.3

, as the output variable. To

test for the existence of a relationship between the input and the

output variables, we have performed experiments with a feed-for-

ward network. Network parameters have been tuned by means of

F-Race (

Birattari, Stützle, Paquete, & Varrentrapp, 2002

). A network

composed of 3 input neurons, 5 hidden neurons and 1 output neu-

ron turned out to be the best possible option for our study.

We started by training the network using for the training set,

180 randomly chosen ﬁrms out of the collected data, and evaluat-

ing the generalisation accuracy of the model on the test set which

is formed by the remaining part of the data (93 ﬁrms). Moreover,

the Neural Network was trained by means of BackPropagation

Momentum (with parameters

= 0.2 and b = 0.5). In order to avoid

over-ﬁtting, we have performed this procedure over 30 different

partitions of data to test the degree of the generalisation of the

results. The goal of this experimental phase was to see whether

Table 2

Main statistics of the three principal co-creation variables.

Component

Mean

STD

Skewness

Kurtosis

Resources and processes

2.290

0.821

0.087

À0.105

Customer relationships

1.857

0.556

0.028

0.251

Mutual learning

5.984

2.457

0.264

À0.126

All

3.973

1.105

À0.078

À0.238

Table 3

Main statistics of the innovation variable.

Innovation metric

Mean

STD

Skewness

Kurtosis

Perception

4.745

1.760

À0.126

À0.286

12722

G. di Tollo et al. / Expert Systems with Applications 39 (2012) 12719–12726

the network is able to correctly generalise the innovation-related

output variable over not-seen-before data. Results for this ap-

proach will be shown on Section

4.1

3.2. Self Organising Map to cluster ﬁrms’ involment in co-creation

values

Data mining tools are very helpful in analysing data, in support-

ing decision and in extracting knowledge from data, but some

unaffordable problems can arise when system responses are not

collected. These problems are tackled by unsupervised learning

methods, that are able to extract key relations among input data

without any support of collected outputs of the system. Many ef-

forts have been concentrated to develop unsupervised cluster algo-

rithms and data reduction algorithms to make relations among

inputs more comprehensible such as k-means clustering (

Mac-

Queen, 1967

), PCA (

Jolliffe, 1986

) or multidimensional scaling

(

Borg & Groenen, 2005

). In our approach we focus on using Koho-

nen Self-Organising Map (SOM) algorithm (

Kohonen, Schroeder, &

Huang, 2001

). A SOM is a type of artiﬁcial Neural Network that is

trained using unsupervised learning to produce a two-dimensional,

discretised representation of the input space of the training sam-

ples, called map. We selected SOM method for its appreciable fea-

tures that allow to recognise groups of similar input vectors

(clustering) even when non-linear relations among variables exist.

Let an initial lattice, the Self-Organising Map, be deﬁned by

= {

, . . . ,

} neurons as an a priori choice, and let be each neuron

associated to a prototype vector m

, formed by a vector of weights

2 R with j = {1, . . . , p} where p = 4 is the number of variables of

the data set matrix (i.e. the three value co-creation components,

deﬁned in Section

2.2

and the perception of innovation, deﬁned

in Section

2.3

). Initially the set of weights is randomly selected in

R. The SOM is iteratively trained, at each iteration t a sample vector

x is randomly chosen from the input data set, euclidean distances

between x and all the m

are then calculated. The neuron

whose

prototype has the closest distance from x

¼ argmin

16i64

ðtÞ À xðtÞk

ð1Þ

identiﬁes the winning neuron at iteration t and it is called Best

Matching Unit (BMU). The weights of the prototype associated to

the BMU and to its closest neighbours in the SOM lattice are ad-

justed towards the x vector. This adjustment decreases with time

and with distance from the BMU according to:

tþ1

¼ m

þ wðx; tÞ Â

ðtÞðx À m

Þ;

ð2Þ

where

(t) is a monotonically decreasing learning coefﬁcient. The

(x, t) function is a gaussian kernel function over the neighbours

of the BMU so that also the neighbouring neurons are moved closer

to the input vector, but with smaller magnitude, at each learning

step. The training procedure is then iterated for all the inputs of

the data set and during the training, data lying near each other in

the input space are mapped onto nearby map units (

Vesanto & Alh-

oniemi, 2000

). The result of the procedure is that the winning neu-

ron is more likely to be a BMU whenever a similar vector is

presented. As more and more inputs are presented, each neuron

in the layer closest to a group of input vectors soon adjusts its

weight vectors toward those input vectors.

4. Experimental results

In this section we are going to show the results obtained by

experimental analysis. Section

4.1

will show the results obtained

using ANN to determine the correlation between co-creation and

innovation; Section

4.2

will show the results obtaining using

SOM to cluster ﬁrms in terms of the degree of their involvement

in co-creation.

4.1. Neural Networks to determine the correlation between co-creation

and innovation

The results obtained by ANN clearly indicate that there is a rela-

tionship between the actual and desired outputs, and this assertion

is of the utmost importance, since it is observed over the test set. It

suggests that, since the network has been trained using the co-cre-

ation component values, the variation of the co-creation compo-

nents is able to explain ﬁrms’ perception of innovation. This

could be seen on

Fig. 1

, where the expected output (innovation)

for the test examples is shown along the x-axis, and the actual net-

work output (innovation) for the same dataset is shown along the

y-axis. Just results obtained by tackling two different partitions of

data are reported. Other partitionings lead to comparable

behaviours.

In order to verify if there is a generalised trend of correlation be-

tween the current network output and the desired output (innova-

tion metric) found over the 30 different training-test partition, we

plot, in

Fig. 2

, the cumulative empirical distribution of Spearman’s

rank based correlation (

Spearman, 1904

) value between desired

and current network’s output values. We decided to introduce a

correlation analysis instead of deﬁning an error measure due to

the lack of such an error measure in previous research. In order

to assess an error measure, we should have introduced a subjective

threshold, without no guarantee on the soundness of this thresh-

old. A correlation analysis instead, just relies on data, without fur-

ther manipulation and without taking into account the variable

scale. Furthermore we decided to use the rank-based correlation

rather than, i.e., Pearson correlation, in order to evidence non-lin-

ear features between variables. It is nonetheless worthwhile to no-

tice that rank based correlation and Pearson indicator lead to

comparable results.

We can see that there exists a positive rank-based correlation

on the variables under examination, and even in the cases where

this relationship appears to be weaker, it is never smaller than

0.5. The correlation measure is greater than 0.85 in 70 % of the

cases, i.e. the positive relationship between variables appears to

be robust. Hence, we can conclude that Neural Networks can be

used to examine the relationship between the co-creation compo-

nent and ﬁrms perception of innovation (see Section

4.1

). This re-

sult is in agreement with the results from the linear regression

analysis provided by

Tanev et al. (2011)

, which shows that there

is a statistically signiﬁcant positive association between the

perception of innovation and the value co-creation components

‘‘Customer

relationships

enabled

through

partnerships

and

cooperation’’ and ‘‘Mutual learning mechanisms’’. The agreement

and the high explanatory power of the linear regression model

(49.0 %, assessed by the adjusted R square value) suggest that

linear models are quite adequate in describing the relationship

between value co-creation and the perception of innovation, also

showing with the additional advantages of being less time

consuming as well as being able to identify the dominant role of

speciﬁc co-creation components. The combination of the results

from the ANN and linear regression analysis provides evidence in

support of one of the initial hypotheses that more co-creative ﬁrms

are in a better position to differentiate themselves by emphasising

the innovative aspects of their new products, processes and

services.

The good agreement between the ANN and linear regression

does not address the question of how good the online innovation

metric is in describing the innovative capacity of ﬁrms. The answer

to this question requires the additional research focusing on the

relationship between the three value co-creation components

G. di Tollo et al. / Expert Systems with Applications 39 (2012) 12719–12726

12723

and some traditional innovation metrics based on the number of

new products, processes and services.

4.2. Results of SOM classiﬁcation

The present section provides the results of the SOM approach

for the classiﬁcation of the ﬁrms in terms of the degree of their

involvement in co-creation. The results provided by the SOM to

the classiﬁcation of ﬁrms in terms of the degree of their involve-

ment in co-creation are compared to previous results (

Tanev

et al., 2011

) generated by K-means cluster analysis.

Both classiﬁcation methods suggest that the ﬁrms can be classi-

ﬁed in four groups.

Table 4

shows the contingency table including

the number of ﬁrms in each of the four groups (clusters) generated

by the two different methods. The table has to be read as follows:

cell (row, col) represents the intersection (in term of ﬁrms) between

the cluster row (produced by k-means) and the cluster col (pro-

duced by SOM).

It can be clearly seen that cluster 3 generated by the k-means

method is entirely contained within one of the groups generated

by the SOM method. The analysis shows that these are exactly

the ﬁrms that are most active in co-creation. The second group of

ﬁrms generated by the SOM method is entirely contained in fourth

group of ﬁrms generated by the K-means method. It could be

pointed out that the good agreement between the two lists of ﬁrms

most active in co-creation is remarkable. It shows that the SOM

method is able to quantitatively identify such ﬁrms opening the

opportunity for the potential application of additional qualitative

comparative analysis to examine their speciﬁc co-creation

strategies.

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

"ris.dat"

0.2

0.4

0.6

0.8

0.2

0.4

0.6

0.8

"newris.dat"

(a)

(b)

Fig. 1. Relationship between desired and actual output for two different train-test partitions. The x-axis corresponds to the expected output value (perception); the y-axis

corresponds to the actual network value.

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

0.5

0.55

0.6

0.65

0.7

0.75

0.8

0.85

0.9

0.95

Rank−Based Correlation Values

Empirical CDF

Fig. 2. Cumulative distribution of the Spearman’s rank based correlation values between desired output (perception) and actual output over 30 train-test partitions.

Table 4

Common elements (ﬁrms) differences amongst clusters found by cluster analysis by

Tanev et al. (2011)

and SOM. CA

indicates the ith cluster obtained by cluster analysis;

SOM

represent the jth clusters obtained by SOM (i, j 2 {1234}).

SOM

105

12724

G. di Tollo et al. / Expert Systems with Applications 39 (2012) 12719–12726

5. Conclusions

The present study provides an ANN analysis to examine the

relationship between the degree of value co-creation activities

and ﬁrms’ innovativeness. Although, it is impossible to claim the

existence of a causal relationship, the results suggest that value

co-creation practices could be considered good indicators of the

ﬁrms innovation-related outcomes such as the degree of online

articulation of the innovative aspects of their new products, pro-

cesses and services. The advantage of such approach can be found

in the opportunity to test the existence of this relationship without

any preliminary assumption about its speciﬁc functional form. This

opportunity appears to be highly relevant given the early stage of

quantitative value co-creation research and the still limited knowl-

edge about the relationship between co-creation and innovation.

In addition, the present work applies SOMs to classify the ﬁrms

in terms of the degree of their involvement in value co-creation.

Therefore, our main contributions of this work should be seen in

the speciﬁc methodological setting, since it could open the way

for applications of ANN modelling to co-creative innovation

research.

We stress out that, up to the authors knowledge, this is the ﬁrst

application of these different Neural Networks models to innova-

tion and co-creation. These two approaches have shown a high de-

gree of ﬂexibility and performance in adaptation, prediction and

classiﬁcation.

We could however suggest as a subject of future research the

development and the comparison of different Neural Networks in

terms of topologies and connections in order to generate reliable

and robust models to predict more complex innovation activities.

One should also compare the SOM approach with other unsuper-

vised appraoches to determine, whether or not, a Neural Network

model platform could be suited to simultaneously model and clas-

sify such kind of data sets. The potential value of a combined appli-

cation of both modelling, ANN and SOM, could be found in their

ability to take into account the inherent complexity and the emerg-

ing nature of value co-creation networks.

We stress out the fact that the results shown here were based

on an online innovation metric that has been recently introduced

in the literature.

Such an approach will provide an opportunity for future re-

search to focus on the development of speciﬁc online innovation

metrics to overcome the limits of more traditional ones, such as

the ones suggested in the OSLO manual. This could open new re-

search areas focusing on the development of business intelligence

and innovation research tools that would increase the utility of

both managers and researchers.

References

Allen, S., Tanev, S., & Bailetti, T. (2009). Components of co-creation, Special issue on

value co-creation. Open Source Business Review Online Journal, November.

Angelini, E., di Tollo, G., & Roli, A. (2008). A neural network approach for credit risk

evaluation. Quarterly Review of Economics and Finance, 48, 733–755.

Birattari, M., Stützle, T., Paquete, L., & Varrentrapp, K. (2002). A racing algorithm for

conﬁguring metaheuristics. In Proceedings of the genetic and evolutionary

computation conference (pp. 11–18). Morgan Kaufmann Publishers.

Bishop, C. (1996). Neural networks for pattern recognition. USA: Oxford University

Press.

Borg, I., & Groenen, P. J. F. (2005). Modern multidimensional scaling: Theory and

applications. Springer.

Bowonder, B., Dambal, A., Kumar, S., & Shirodkar, A. (2010). Innovation strategies for

creating competitive advantage. Research-technology Management, 53(3), 19–32.

Chesbrough, H. W. (2011). Open services innovation – Rethinking your business to

grow and compete in a new era. San Francisco: Jossey-Bass, pp. 53–67 [Chapter 3:

Co-create with your customers].

Christensen, C. M. (2006). The Ongoing process of building a theory of disruption.

Journal of Product Innovation Management, 23, 39–55.

Desai, D. (2010). Co-creating learning: Insights from complexity theory. The

Learning Organisation, 17(5), 388–403.

Droge, C., Stanko, M., & Pollitte, W. (2010). Lead users and early adopters on the

web: The role of new technology product blogs. Journal of Product Innovation

Management, 27, 66–82.

Etgar, M. (2006). Co-production of services. In R. Lusch & S. Vargo (Eds.), The Service

Dominant Logic of Marketing. Armonk, NY: M.E. Sharpe Inc.

Etgar, M. (2008). A descriptive model of the consumer co-production process.

Journal of the Academy of Marketing Science, 36(1), 97–108.

Ferrier, W. (2001). Navigating the competitive landscape: the drivers and

consequences of competitive aggressiveness. Academy of Management Journal,

44(4), 858–877.

Hagan, M., Demuth, H., & Beale, M. (1996). Neural network design. Colorado: PWS

Pub.

Hicks, D., Libaers, D., Porter, L., & Schoeneck, D. (2006). Identiﬁcation of the

technology commercialisation strategies of high-tech small ﬁrms. Small

Business Research Summary, December.

rs289tot.pdf>

Hinton, G., & Sejnowski, T. J. R. (Eds.). (1999). Unsupervised learning: Foundations of

neural computation. The MIT Press.

Hykin, S. (1999). Neural networks: A comprehensive foundation (Second ed.). Prentice

Hall International, Inc.

Jolliffe, I. (1986). Principal component analysis. New York, NY: Springer.

Kohonen, T., Schroeder, M. R., & Huang, T. S. (Eds.). (2001). Self-organizing maps.

Syracus, NJ, USA: Springer-Verlag New York, Inc.

Kristenson, P., Matthing, J., & Johansson, N. (2008). Key strategies for the successful

involvement of customers in the co-creation of new technology-based services.

International Journal of Service Industry Management, 19(4), 474–491.

Lusch, R., & Vargo, S. (Eds.). (2006). The service dominant logic of marketing. Part III:

Co-production, collaboration, and other value-creating processes (pp. 105–179).

New York.

MacQueen, J. B. (1967). Some methods for classiﬁcation and analysis of multivariate

observations. In L. M. L. Cam & J. Neyman (Eds.). Proceedings of the ﬁfth Berkeley

symposium on mathematical statistics and probability (vol. 1, pp. 281–297).

University of California Press.

Michel, S., Brown, S., & Gallan, A. (2008). An expanded and strategic view on

discontinuous innovations: Deploying a service-dominant logic of marketing.

Journal of the Academy of Marketing Science, 36(1), 54–66.

Midgley, D. (2009a). Co-creating the innovation with customers. In The innovation

manual – Integrating strategies and practical tools for bringing value innovation to

the market (pp. 143–190). Chichester, UK: John Wiley and Sons.

Mindgley, D. (2009b). The Innovation Manual (vol. 5). Chichester: John Wiley and

sons, pp. 143–190 [Chapter 5: Co-creating the innovation with customers].

Nambisan, S. (2009). Virtual customer environments: IT-enabled customer co-

innovation and value co-creation. Information Technology and Product

Development, Annals of Information Systems, 5(2), 127.

Nambisan, S., & Baron, A. (2009). Virtual customer environments: Testing a model of

voluntary participation in value co-creation activities. Journal of Product

Innovation Management, 26, 388–406.

Nambisan, S., & Nambisan, P. (2008). How to proﬁt from a better ’virtual customer

environment. MIT Sloan Management Review, 49(3), 53–61.

Ng, I. (2010). The future of pricing and revenue models. Journal of Revenue and

Pricing Management, 9(3), 276–281.

Payne, A., Storbacka, K., & Frow, P. (2008). Managing the co-creation of value.

Journal of the Academy of Marketing Science, 36, 83–96.

Prahalad, C. K., & Krishnan, M. S. (2008). The new age of innovation. New York:

McGraw Hill.

Prahalad, C. K., & Ramaswamy, V. (2003). The new frontier of experience innovation.

MIT Sloan Management Review, 44(4), 12–18.

Prahalad, C. K., & Ramaswamy, V. (2004). The future of competition – Co-creating

unique value with customers. Boston: Harvard Business School Press.

Ramaswamy, V., & Gouillart, F. (2010a). Building the co-creative enterprise. Harvard

Business Review, 100–109.

Ramaswamy, V., & Gouillart, F. (2010b). The power of co-creation. New York: Free

Press.

Reed, R. D., & Marks, R. J. (1999). Neural smithing: Supervised learning in feedforward

artiﬁcial neural networks. Cambridge, MA: The MIT Press.

Roberts, D., Bake, S., & Walker, D. (2005). Can we learn together? Co-creating with

customers. International Journal of Market Research, 47(4), 407–427.

Romero, D., & Molina, A. (2009). Value co-creation and co-innovation: Linking

networked organisations and customer communities. Leveraging knowledge for

innovation in collaborative networks. IFIP Advances in Information and

Communication Technology, 307, 401–412.

Rumelhart, D. E., Hinton, G. E., & Williams, R. J. (1986). Learning internal

representations by error propagation. In D. E. Rumelhart et al. (Eds.), Parallel

distributed processing (pp. 318–362). Cambridge: MIT Press.

Sawhney, M., Gianmario, V., & Prandelli, E. (2005). Collaborating to create: The

internet as platform for customer engagement in product innovation. Journal of

Interactive Marketing, 19(4), 4–17.

Spearman, C. (1904). The proof and measurement of association between two

things. American Journal of Psychology, 15, 72–101.

Tanev, S., Bailetti, T., Allen, S., Milyakov, H., Durchev, P., & Ruskov, P. (2011). How do

value co-creation activities relate to the perception of ﬁrms’ innovativeness?

Journal of Innovation Economics, 1(7), 131–159.

Tanev, S., Knudsen, M., & Gerstlberger, W. (2009). Value co-creation as part of an

integrative vision for innovation management. Special issue on value co-

G. di Tollo et al. / Expert Systems with Applications 39 (2012) 12719–12726

12725

creation. Open Source Business Review Online Journal, December.

www.osbr.ca/ojs/index.php/osbr/article/view/1014/975>

Vargo, S. L., & Lusch, R. F. (2004). Evolving to a new dominant logic for marketing.

Journal of Marketing, 68(January), 1–17.

Vesanto, J., & Alhoniemi, E. (2000). Clustering of the self-organizing map. IEEE

Transactions on Neural Networks, 11(3), :586–600.

von Hippel, E. (2001). Perspective: user toolkits for innovation. Journal of Product

Innovation Management, 18, 247–257.

von Hippel, E. (2006a). Why many users want custom products. In Democratization

of innovation (pp. 33–43). Cambridge: MIT Press.

von Hippel, E. (2006b). Application: Searching for lead user innovations. In

Democratization of innovation (pp. 133–146). Cambridge: MIT Press.

Zemella, G., De March, D., Borrotti, M., & Poli, I. (2011). Optimised design of energy

efﬁcient building façades via evolutionary neural networks. Energy and

Buildings, 43(12), 3297–3302.

http://dx.doi.org/10.1016/j.enbuild.2011.10.006.

ISSN 0378-7788.

12726

G. di Tollo et al. / Expert Systems with Applications 39 (2012) 12719–12726

Document Outline

Neural Networks to model the innovativeness perception of co-creative firms
- 1 Introduction
- 2 Value co-creation, innovation, and complexity
  - 2.1 Research methodology
  - 2.2 Value of co-creation components
  - 2.3 Online innovation metric
- 3 Statistical approaches for a quantitive analysis of innovation and co-creation
  - 3.1 A Neural Network approach to model innovation based outcomes
  - 3.2 Self Organising Map to cluster firms’ involment in co-creation values
- 4 Experimental results
  - 4.1 Neural Networks to determine the correlation between co-creation and innovation
  - 4.2 Results of SOM classification
- 5 Conclusions
- References

Download 194.4 Kb.

Do'stlaringiz bilan baham: