Sie sind auf Seite 1von 48

MEASURING INNOVATION

MANAGEMENT
3.1 Introduction

Measuring of Innovation in this study was done by identifying


success factors of innovation projects from various literature (refer
chapter2) and their importance and satisfaction scores were obtained by
administering the questionnaire among the business users, management
personnel and technical personnel involved in the design and
implementation of Information Technology Service projects.

The data collected from the business users, technical personnel


and management personnel for the study was processed by using SPSS
20.0. The statistical results are given in Tables l?om 3.1 to 3.27.

The demographics of respondents are portrayed in the following


section.

3.2 Demographic characteristics of the respondents

The demographic characteristics of the respondents in


Information Technology Services are presented in Table 3.1 to 3.4.

A total of 3000 responses for 100 projects are received, while


studying the demographic value it is identified that a total of 2579 Males
and 421 Females have take part in the providing the responses are
depicted in Table 3.1 .

The average age of respondents for 100 prajccts is ranging


between 25.2 years to 39.6 years. The average age of 3000 users is
31.94 years. Table 3.2 shows the average age of all the respondents in
100 projects.
Table 3.1 Project Wise Men and Women respondents

The average experience of all the response is 8.68 years as the


Information Technology Industry encourages young blood. There is not
much significant difference in the average experience between men
(8.68 years) and women (8.66 years). Minimum experience in a project
is 4.9 years when compared to a maximum of 12.6 years. Table 3.3
11A
depicts the average experience of employees in Information Technology
Projects.

Table 3.2 Project Wise Average age of people invotved in ITS


Table 3.3 Project wise average experience of respondents
Proj.No I Exp

Table 3.4 CMM Level wise average age, average experience and number
of men and women employees
CMM Level wise average age, average experience and number of
Males and Females are depicted in Table 3.4. Maximum numbers of
men are working in CMM level 2 projects and women in CMM level 4
projects. The maximum average experience is shown in CMM level 4
projects whereas the minimum experience in CMM level 1. Maximum
average age of personnel are involved in CMM level 4 projects whereas
minimum is from CMM level 1 projects.

The Normalized Means of factors of success with respect to


business users, management personnel and Technical personnel for each
project are given in Table 3.5.

Means are calculated using the formula given below.


n
Mean = c xi/n

Where xi = intensity score perceived by the business users


involved in project.
i = 1 to n = number of factors considered for assessing success
rate by the business users (i.e. 17 factors)

Similarly management personnel (17 factors) and technical


personnel (17 factors) perceptions about success factors were computed
and shown in Table 3.5.

Normalized Weightage is calculated by multiplying intensity


with importance score perceived by all respondents for each project i.e.
business users, management personnel, and technical personnel
separately on the total scale of 100 and the data is shown in Table 3.6.
Table 3.5 Normalized Means of Intensity of success factors of Business
Users, Management Personnel and Technical Personnel
3.3 Project wise computations

Rate of satisfaction is measured on intensity scale of 1 to 5 for


the factors considered for business users, management personnel and
technical personnel each and Importance is measured on importance
scale of 100 for all factors together to each type of personnel.

The rate of satisfaction as perceived by business users,


management personnel and technical personnel were presented in Table
3.5 for 100 Projects

Total weight of success for project 1 is calculated by summing


the Weightage scores of business users, management personnel and
technical personnel. For Example Project 1 Total Weightage = 481.65 1-
439.85 + 41 1.24 = 1333. Table 3.6 gives for all 100 projects.

The project which secured 85% success Weightage are


considered as successful whereas others are treated as unsuccessful.
That means if the collective Weightage of business users, management
personnel, and technical personnel weights of success are greater than or
equal to 1275 then the project is considered as successfbl, if total is less
than 1275 then the project is considered as unsuccessful. In the above
example project 1 is successful as the total Weightage is 1333.

Table 3.7, Table 3.8 and Table 3.9 shows computations for
project 1 of business users, management personnel and technical
personnel.

Similarly for all 100 projects Normalized Weightage success


scores are calculated and presented in Tables listed in Annexure E, F
and G.
Table 3.7 Project 1 computation for Business Users

Table 3.8 Project 1 computation for Management Personnel


3.4 Exploration of Critical Success Factors

To explore the underlying factors leading to the success of


Innovation Projects are being identified by applying the Factor Analysis
on Business User factors (1 7 items), Management Personnel factors ( I 7
items) and Technical Personnel factors (17 items) and are depicted in
Table 3.10

To extract the factors, Principal Component Analysis and


Varimax with Kaiser Normalization rotation analysis has been applied
using SPSS and the results have been given in the Tables 3.1 1 to 3.13

From data of Table 3.1 1 critical success factors as perceived by


business users for success of innovation projects are Business
Enhancement, Risk Improvisization, Market edge on Innovation, Vision
Enrichment, and Reengineering, with factor loadings higher than 0.597,
with reliability greater than 0.43 1.
143
From data of Table 3.12 critical success factors identified by the
management personnel for success of innovation projects are Flexibility
to Change, Knowledge Sharing, Value added project, Resourcing,
Individual Excellence with reliability greater than 0.5 and Process
Oriented identified with factor loading of 0.491, with reliability of
0.385.

Table 3.10 Underlying factors leading to the success of Innovation in


Information Technology Projects
Management
S.No Business Users Personnel -
1 Supports value creation and Peer support across
addition boundaries -
2 Maintenance of dignity of Readiness for regular Review of Functional
workforce change Specifications
3 Designing of decentralized well defined flexible Review of Technical
small business solutions execution process Specifications
4 Explores news ideas Management Vision Employees Motivation
Helps in magnification of Providing support lo Directed functional
sales colleagues Specifications -
6 Creates boundary less Partner's consultation Optimal use of
knowledge sharing for open innovation. Technology

7 Enhances competition in Effective decision Clarity of' technical


targeted market making specifications
Glorifies effective decision
making Rjskva!u@&n- . .- Acce~tanc=ocedu'_e_~~
Relative price and cog - - Consultation of technical ~ u i functional
i
reduction ,FEE?E?l - dedicated teams
.. .. ~
10 .- --.- in absurdity.
See value Business
. results
--. .impact
- Knowledge base -
, High potential for risk
absorption
Intranet as a
communication agent Exploration thinking
Creates positive impact on Project teams say in Encouragement of Idea
12
brand imaging resourcing generation
Helps in vision Implementation after
l3 work-life balance trails
~mprovisization

14 Creates Learning environment Management confidence Resource skill


from others mistakes on technical resources enhancement
Management
15 Ease of implementation in Involvement in users Reduction of cycle or
1 existing facilities
requirements execution time
Are standards of
16 Projects are well defined with Strategic Change
flexible execution process Management - performance defined
17 Edge on the competitors Effective line of control Experimentation
Table 3-11Critical Success factors of innovation in ITES projects in view
of business users
Cmnbacb's
Factor Loading Mean Alpha
Coceient ,

Factor 1- Business Enhancement


S~pportsvalue creation and addition 4.94
970
(1.235)
Maintenance of dignity of workforce 4.94
'967
(*.238J-)_
Designing of decentralized small business 4.85
solutions ,908
-- -.L i'499).- ...... 4.883
,
,879
Explores news ideas anytime

Helps in magnification of sales


.+....... ..
900
.___--..- *.=
4.93
1.-
4.77
..

(f.419)
Creates boundary less knowledge sharing 4.87
s97
(f.338)
Factor2 -Market edge on Innovation
4.96
Enhances competition in targeted market 9M
(1.190)
4'96
Glorifies effective decision making '944 4.948 .43 1
-" .- . .- -. . >..20S).
-. ..-.
4.9 1
Relative price and cost reduction .724
-
- .. 11.293) - .... .- ...... .
Factor3 - Risk Improvisization
4.58
See value in absurdity .us7
(*'601) 4.61 8 ,759
4.66
High potential for risk absorption 843
(k.475)
Factor4 - Vision Enrichment
4.65
Creates positive impact on brand imaging 87"&.47~)
4.72
Helps in vision Improvisization "4
( t . 4 5 3 L 4,425 -.468
C:reates Learning environment from others 3.91
mistakes -'621 (k.616)
--.
Factor5 - Reengineering , ,,, , .. ... .- _. . . --
4.72
Ease of implementation in existing facilities .799
.-..A.
(*'448)
."--.--..-" 1
".

4.825 .545
4.93
It helps have an edge on the competitors
-
"
77"1.26~)
Factor6 - Execution excellence 1
Projects are well defined with flexible 4.91
I'93 7 , (1.292)
execution process
Table 3.12 Critical Success factors of innovation in lTES projects in view
of Management Personnel

K k 7 i f e balance 4.46
.R75
(A.555)
Management confidence'on technical 4.43
resources '726
(k.629) 4.432 .735
Management Involvement in users 4.41
requirements '548
(L.870)

-
~acto;6 Process Oriented
4.47
--
Strategic Change Managemenl 76R
(k.617) -
- ,
4.43 1 ,385
Effective line of control 4.40
.7'7
.
...
-
(1.719)
Table 3.13 Critical Success factors of innovation in ITES projects in view
of Technical Personnel

From data of Table 3.13 critical success factors according to thc


technical personnel for success of innovation projects art: Technical
Enhancement, Innovation Culture and Experimentation identified with
factor loading higher than 0.58, with reliability greater than 0.858.
From the above three tables it can be concluded that there is a
significant difference in the critical success factors of innovation of
projects for business users, management personnel and design
personnel.

Table 3.14 Means of Intensity of success factors for 100 Projects


Std. 95% Conndence
$.No N Mean Deviat
Error
Stdm
Interval for ern Mini Mlrxl
ion mum mum
Lower Upper
Bound Bound -
1 100 4.7464 ,07391 .00739 4.7317 4.761 1 4.48 4.83

2 100 4.2335 .33383 ,03338 4.1672 4.2997 3.56 4.62

3 100 4.1762 ,20486 ,02049 4.1356 4.2169 3.78 4.56

Total 300 4.3854 .34430 ,01988 4.3462 4.4245 3.56 4.83

1- Intensity of success factors as perce~vedby Busmess Users


2- Intensity of success factors as perce~vedby Management Personnel
3- lntensity of success factors as perceived by Techn~calPersonnel
N- Number of projects -

Table 3.15 ANOVA Results of Mean differences among business


users, management personnel and technical personnel
-3

Sum of
Square Mean
s df Square F Sip.
Between Groups 19.716 2 9.858 186.140 .000
-. -~- ..
W ~ t h i nGroups 15 729 297 .053
- .-
Total 35 444 299
, -__
_ _ -_ _ __".

Table 3.14 portrays Means of Intensity of success factors of


projects perceived by business user, management personnel and
technical personnel factors.

From the ANOVA Table 3.15 it is understood that percevance of


intensity of success factors by business users, management personnel
and technical personnel are significantly differing at 0.05 levels. Thus
Hypothesis 1 : ''Intensity of dimensions of success factors is di#crent
for business users, management personnel and technical personnd' is
acceptable.

3.5 Modelling Innovation Project Success using Neural Networks

In order to find the success rate of innovative projects,


Weightage's obtained collectively from factors of success of business
users, management personnel and technical personnel(Tab1e 3.10) with
age and experience as covariates and project success indicator as
dependent variable are inputted to the multilayer Perceptron Neural
Network model and the model is tested with the help of SPSS 20.0
version.

Multilayer Perceptron

The Multilayer Perceptron (MLP) procedure produces a


predictive model for one or more dependent (target) variables based on
values of the predictor variables. Following below arc the sequence of
steps that are executed with multilayer Perceptron option.

Stepl: Setting the random seed (Fig 3 . 1 ) allows replicating the


analysis exactly. To set the random seed, fiom the menus choose:
Transform > Random Number Generators

Step2: Random number generation: Select Set Starting Point.


Select Fixed Value, and type 9191972 as the value. Click OK as shown
in Fig 3.2

Step3: Creation ofpartition variable: Approximately 80% of the


Innovative Information Technology Projects were randomly assigned to
149
the training sample and 20% to a holdout sample. A partition variable
will be necessary to exactly recreate the samples used in those analyses.

To create the partition variable, from the menus choose:


Transform > Compute Variable as shown in Fig 3.3. Type partition in
the Target Variable text box. Type 2*rv.bernoulli(0.8)-1 in the Numeric
Expression text box.

This sets the values of partition to be randomly generated


Bernoulli variates with a probability parameter of 0.8, modified so that it
takes values 1 or -1, instead of 1 or 0. Recall that cases with positive
values on the partition variable are assigned to the training sample, cases
with negative values are assigned to the holdout sample, and cases with
a value of 0 are assigned to the testing sample.

Step4: Compute variable: Click OK in the Compute Variable


dialog box as depicted in Fig 3.4. Approximately 80% of the Innovative
Information Technology Projects previously will have a partition value
of I . These Innovative IT Projects will be used to create the model. The
remaining Innovative Information Technology Projects will have a
partition value of -1.

Steps: Running Multilayer Perceptron Analysis: To run a


Multilayer Perceptron analysis, from the menus choose: Analyze >
Neural Networks > Multilayer Perceptron as portrayed in Fig 3.5 and
Fig 3.6. Select Project Success Indicator as a dependent variable. Select
Normalized Means of Business User, Management Personnel, and
Technical Personnel as a factor. Select Age and Experience in years as
covariates. Click the Partitions tab.
Step6: Selection of Dependent Variable, Factors and Covariate
is depicted in Fig 3.7

Step 7: Using partition variable to assign cases: Fig 3.8 and 3.9
portrays setting of partition variable for the multilayer Perceptron
process. Select Use partitioning variable to assign cases and then select
partition as the partitioning variable.

Step8: Setting output window options: Click on the Output tab of


Multilayer Perceptron window dialog box. Deselect Diagram in the
Network Structure group. Select Cumulative gains chart check box as
shown in Fig 3.10. The Residual by predicted chart is unavailable
because the dependent variable is not scale. Select Independent variable
importance analysis and click OK button.

These selections generate the following command syntax:

*Multilayer Perceptron Network.


GET DATA
/TYPE=XLS
/FILE='C:\Documents and
Settings\Administrator\Desktop\Pinakapani\Master-Data-BU-MP-TP-
7-Mar.xlsl
/SHEET=name 'Master Data'
/CELLRANGE-full
/READNAMES=on
/ASSUMEDSTKWLDTH=32767.
EXECUTE.
DATASET NAME DataSetl WINDOW=FRONT.
SET SEED=9191927.
COMPUTE p a r t i t i o n = Z * r v . b e r r i o u L J . i (0.8)-1.
EXECUTE.
*Multilayer Perceptron Network.
MLP Project.Success1nd (MLEVEL=N) BY GMBU GMMP FMTP WITH Age Exp
/RESCALE COVARIATE=ADJNORMALIZED
/PARTITION VARIABLE=partition
/ARCHITECTURE AUTOMATIC=YES (MINUNITS=l MAXUNITS-50)
/CRITERIA TRAINING=BATCH OPTIMIZATION=SCALEDCONJUGATE
LAMBDAINITIAL-0.0000005 S~GMAINITIAL=0.00005 INTERVALCENTER=O
INTERVALOFFSET=0.5 MEMSIZE=1000
/PRINT CPS NETWORKINFO SUMMARY CLASSIFICATION IMPORTANCE
/PLOT NETWORK ROC GAIN LIFT PREDICTED
/STOPPINGRULES ERRORSTEPS= 1 (DATA-AUTO) TRAININGTIMERZON
(MAXTIME=15) MAXEPOCHS-AUTO ERRORCHANGE-1.OE-4
ERRORRATIO=0.0010
/MISSING USERMISSING=EXCLUDE.

The procedure builds a network for the nominal-level variable


default, based upon the factors of business users, management personnel
and technical personnel and covariates as age and experience.

Fig 3.1 Setting the random seed

Recode lnto Same Var~abler.. .


Recode into Different Variables.. .
Automatic Recode...
V~sualBmning,,,
Opbmd B~nnlng

Rank Cases.. .
Date and Tvno Waard.. .
Create Tlme Series..
Replace Missing Values.. .
Random Number Generat

Fig 3.2 Random Number Generator Screen

Cwrent Active Generator SPSS '12CDmpatlbSc

w The a c t w pneratw settho appYes mmuimteiy and to future


-
s
o
.

Some procedures h a w 6 mlernal random number generaton Ssc Me


help for a mcnplete Id.
Fig 3.3 Creation of partition variable
Ele Edit Mew D a b
.-- ."
hdd-ens Window &dp

Count Values within Cases,,,

Recode into Same Variables,, a

Recode into Different Variables,..


hutomatr Recode, ,,
ri~sualBnn~ng,, ,
Optlmal Bmnlng.. .

Rank Cases.. ,

Date and T~meWizard, ,


Create T~meSeries,, ,
Replace Mlss~ngValues,,,
Random Number Generators., ,

Fig 3.4 Compute variable dialog box

=::-1-
Target Variable
;-3.
N u r n ~ l cExpression

2*rv brnrnoull1(071.1

i
d Level of educdlon [cd]
@ Years wln rurrcnl em
@ Years at current addre ArYhrndlc
/ Household Income In th CDF 8 Noncentral CDF
/ Debt to ("come ratlo (x Conversion
.@ Crcdn card debt m tho Current DatslT~ma
@ Other debt m thousand
&I Prev~ouslyd p f s ~ ~ l eIdd
@ Predlded dalauk, mod Fundom and S~ecm\
Varm
8 Predicted deiaui? mod
@ PrCdlCtrd deiaui?,mod

.. .

. .
Fig 3.5 To get into Multilayer Perceptron dialog box

1 @neral Linear Model


'I
I Mi@ Models
'I

~oonpsrametrlcTeds b

Forecasting b
Survival P
~~issrng
v3ue Analysts.
I M#~plsResponse '1
Complex Samples
Quality Control

Fig 3.6 Multilayer Perceptron dialog box


Fig 3.7 Selection of Dependent Variable, Factors and
Covariates

TO cnanae Inc mcasurcmem lcvcl or a varmas. rlgm-cllcu


the VDlrlablc in the Vwiabbs Ust

Fig 3.8 Using partition variable to assign cases

~ e ? b o nD a t s z r l

@Randomly assgn carna based on r n ~ ~ numbers


v s ot c e ~ s s

CMM Lave1 l C M M eve11

, ", .. .
www -

155
The case processing summary Table 3.16 shows that 69 cases
were assigned to the training sample and 1 1 to the holdout sample. The
20 cases excluded from the analysis.

The network information Table 3.17 displays information about


the neural network and is usefbl for ensuring that the specifications are
correct and inferences are as follows:

*:* The number of units in the input layer is the number of covariates
plus the total number of factor levels; a separate unit is created
for each category of business users, management personnel and
technical personnel, none of the categories are considered
"redundant" units as is typical in many modelling procedures.

9 A separate output unit is created for each category for a total of


two units in the output layer.

*:* Automatic architecture selection has chosen eight units in the


hidden layer.

The model summary Table 3.18 explains that information about


the results of training and applying the final network to the holdout
sample.

*:* Cross entropy error is displayed because the output layer uses the
softlnax activation function. This is the error function that the
network tries to ~ninimizeduring training.

*:* The percentage of incorrect predictions is taken from the


classification table.

*:* The estimation algorithm stopped because the rnaxillluln number


of epochs was reached. Ideally, training should stop because the
error has converged. This raises questions about whether
something went wrong during training and is something to keep
in mind while further inspecting the output.

Table 3.16 Case Processing Summary


N Percent

Training 69 86.2%
Sample
oldo out 11 13.8%

Valid 80 100.0%

Excluded 20

Total 100

Table 3.17 Network Information


1 GM BU

Factors 2 GM MP

3 GM TP

[nput Layer 1 Age


Covariates
2 E ~ P

escaling Method for Covariates bdjusted normalized

umber of Hidden Layers 1


I
umber of Units in Hidden Layer 1 a 8
I

ctivation Function yperbolic tangent


I I

ependent Variables 1 roject Success Indicator


--L.- ---
Number of Units 2
Output Layer ,
Activation Function Softmax
I

rror Function ross-entropy


I I
a. Excluding the bias unit
Table 3.18 Model Summary - Error Ratio
Cross Entropy Error .011
Percent Incorrect Predictions 0.0%
Training Training errar ratio
Stopping Rule Used criterion (.001)
achieved
Training Time 0:OO:OO.OS
Holdout Percent Incorrect Predictions 18.2%

Dependent Variable: Project Success Indicator

Table 3.19 Model Summary Predicted -

Overall Percent 0.0% 100.0% 81.8%


. --
Dependent Variable: Project Success Indicator (l-
Success, O-Failure)

The classification Table 3.19 shows the practical results of using


the network. For each case, the predicted response is Yes if that case's
predicted pseudo-probability is greater than 0.5. For each sample:

*:* Cells on the diagonal of the cross-classification of cases are


correct predictions.

*:* Cells off the diagonal of the cross-classification of cases are


incorrect predictions.
Of the cases used to create the model, 65 Innovative IT projects
are classified as successfU1. 4 innovative IT Projects are classified as not
successful. Overall, 100% of the training cases are classified correctly,
corresponding to the 0% incorrect shown in the model summary Table
3.19.

Classifications based upon the cases used to create the model


tend to be too "optimistic" in the sense that their classification rate is
inflated. The holdout sample helps to validate the model; here 100% of
these cases were correctly classified by the model. This suggests that,
overall; this model is in fact correct about four out of four times. So it
can be concluded that Neural Network model is useful for finding the
success rate of innovative Information Technology Service Enterprise
projects. Thus Hypothesis 2: "The success of innovation in
Information Technology Projects can be modelled with technique of
fuzzy logic using neural networks" is accepted.

Validation of the Model

Combined with the stopping rule reported in the model summary


Table 3.1 6, this makes to suspect that the network may be over trained;
that is, it is chasing spurious patterns that appear in the training data by
random variation.

SpecifL a testing sample to help keep the network "on track." The
model created the partition variable so that it would exactly recreate the
training and holdout samples used in the logistic regression analysis;
however, logistic regression has no concept of a "testing" sample. Let's
take a portion of the training sample and reassign it to a testing sample.
Following below is the step by step procedure to validate the model:
Creating the Testing Sample

Stepl: Recalling the compute variable dialog box: Recall the


-
Compute Variable dialog box. Type partition rv.bernoulli(0.2) in the
Numeric Expression text box as shown in Fig 3.1 1.

Step2: Working with IF statement: Click on if statement, Select


Include if case satisfies condition. Type partitiom0 in the text box.
Click Continue. The steps are show in Fig 3.12 to 3.14.

Step3: Resetting the value of partition variable: Click OK in the


Compute Variable dialog box to reset the value of partition variable as
shown in Fig 3.15.

This resets the values of partition that were greater than 0 so that
approximately 20% take the value 0, and 80% remain valued at 1.
Overall, approximately 100*(0.7*0.8)=56% of the projects will be in the
training sample, and 14% will be in the testing sample. Projects which
were originally assigned to the holdout sample remain there.

Step4: Running the analysis: Recall the Multilayer Perceptron


dialog box and click the Save tab. Select Save predicted pseudo-
probability for each dependent variable as shown in Fig 3.16. Click OK,
this will trigger the analysis and the reports will be generated.

Table 3.20 shows that of the 69 cases originally assigned to the


training sample, 3 have been reassigned to the testing sample.
The only change to the network information Table 3.21 is that the
automatic architecture selection has chosen nine units in the hidden
layer.

The model summary in table 3.22 shows a couple of positive


signs: The percentage of incorrect predictions is roughly equal across
training, testing, and holdout samples.

The classification Table 3.23 shows that, using 0.5 as the pseudo-
probability cut-off for classification, the network docs considerably
better at predicting successful Information Technology Projects than un-
successful.

Fig 3.11 Recalling the compute variable dialog box

Mma~cExpression
.",",.,. . -- ". ."..
artton. rv barnnrlll(0 211

Proled Success lrld IP


4Total Projcd SS ITdalP
8 GM BU lGMHUl
4GM MP ICMMP] CW 8 NuncentralCDF
4 OM TP [CMTPj
8 BUSS [BUSS] Current Dalelllme
4 W 55 [MPSS] Date Ardhrnetlc
4 TP SS [TPSS] Dnte Creation
&*CMM Level [CMMLevell "

B Age / U ? C ! ~and
..~ Special
I ~ Vu!nble!
8 EXP
fer ale
8-Female
Fig 3.12 Clicking on If case

Fig 3.13 Selecting include if case satisfies condition

&'& 0 Include $1 cases

-lG"-1 ---
49 h i * L z r i
&pro)& ~ u c r c s smu [P
P Total Prolea SSfTotaR
l ~ b d ife k~e;atml~~&IHl --:
] ,,,,,,

@ OM BU [GMEU]
# GM MP IGMMPI
# OM TP lOMTPl
8'suss IBUSSl
bS E-- -
.
".
'Al
. .
Funalan gaup
-
- 1
@ MP 5 5 IMPS51 ~~~~&~
aQu&ua
IAr~hMlc
4 TP 55 I l P b i l CDF 8 Nanccnlrsl CDF
* CMM Lwei lLMMLevel1 Conversl~n
# Age
@E ~ P
4". Ma16
am& Date Arthmetc

& Fcmalr
# partnlon m@h p~ncttonsand Spclai Varlablcs

-- . -

-.--....
Fig 3.14 Setting include if case satisfies condition

Fig 3.15 Resetting the values of the partition variable


Fig 3.16 Save dialog box
- -

save ersonaa vMua or category for each aePendcnt variable


H l S m preooted
~ peeud. plDDDLlllRy tor aaon aapararn varlmwa

lhlnn~r.ol i l n \ * t ~\rHvlnolm
(*i &tomdic.lly amorma unlgur nanaa
s s l s n tws omlon n you want tn add a new ast ol saved varlabler to your d d a s d oach tmr you run a model

,.i cuaom Mmes


spcillyname3 for tha varlab~rsI? you j a l r d thin nplmn sly c ~ ~ s i mvmriablsa
g wah the sams namn or rnnt nemn are
rewoad sscn t)ma you run a modol

Table 3.20 Case Processing - Training


Sample

I -
N 1 ~ e l w n l ]

1 _-.I
Valid

I Excluded 13 ~ - I
-
Table 3.21 Network Information Training Sample
1 GM BU
-
Factors 2 GM MP

3 GM TP

Input Layer 1 Age


Covariates
2 E ~ P

Number of Units" 93

Rescaling Method for Covariates Adjusted normalized

Table 3.22 Training Model Summary


/Cross Entropy Error ,096
i
ercent Incorrect Predictions I 0.0%
Training 1 consecutive step(s)
Stopping Rule Used with no decrease in error"

Training Time 0:OO:OO.OS

Cross Entropy Error .002


Testing -
kercent Incorrect Predictions 0.0%

Holdout ercent Incorrect Predictions 11.1%


L - _ I -
Dependent Variable: Project Success Indicator
a. Error computations are based on the testing sample.
Table 3.23 Classification- Training Sample

Cumulative Gains and Liff Charts

The cumulative gains chart shows the percentage of the overall


number of cases in a given category "gained" by targeting a percentage
of the total number of cases. Top 20% would contain approximately
50% of the Information Technology Projects that are not successful; the
top 30% of cases would contain 70% of unsuccessful Information
Technology Projects, and so on. 100% of the scored dataset will obtain
all of the Information Technology Projects that are unsuccessfbl in the
dataset.

The diagonal line is the "baseline" curve; observing at 10% of the


cases from the scored dataset at random, the expected "gain" would be
approximately 10% of all of the cases that actually take the category
Yes(1). The farther above the baseline a curve lies, the greater the gain.
You can use the cumulative gains chart to help choose a classification
cut-off by choosing a percentage that corresponds to a desirable gain,
and then mapping that percentage to the appropriate cut-off value.

"Desirable gain" depends on the cost of Type I and Type I1


errors. Type I error is the cost of classifying successful Information
Technology projects as unsuccessful. Type I1 error is the cost of
classieing unsuccessful Information Technology projects as successful.
If critical factors are of primary concern then Type I errors should be
lowered. On the cumulative gain charts Fig 3.17 might correspond to
rejecting success in Information Technology projects enterprises in the
top 40% of pseudo-predicated probability of '1" which captures nearly
90% of the possible unsuccessful Information Technology projects but
removes nearly half of the factors.

If growing project success base is the priority, then lower should


be Type I1 error. On the chart, this might correspond to rejecting the top
lo%, which captures 30% of the un-successful Information Technology
Projects and leaves most of the factors intact. Usually, both are major
concerns, so using decision rule helps for classifying factors that gives
the best mix of sensitivity and specificity.

The lift chart Fig 3.18 is derived fi-om the cumulative gains chart;
the values on the y axis correspond to the ratio of the cumulative gain
for each curve to the baseline. Thus, the lift at 10% for the category
Yes(1) is 30%/10% = 3.0. It provides another way of looking at the
information in the cumulative gains chart. The cumulative gains and lift
charts are based on the combined training and testing samples.
Fig 3.17 Gain Chart
- - - - -

70%"

60%-

0% 10% 20% 30% 40% SOU 60% 70% 80% SO% 100%

Percentage

Dependent Var~able.Project Success Ind

Fig 3.18 Lift Chart


0
1

Percentage
Dependent Variable: Project Success Ind
3.6 Impact of CMM level on success rate of projects

In this section various characteristics that are derived for the 51


factors of business users, management personnel and technical
personnel are discussed.

Business Users Perspective:

Fig 3.19, it can be understood about the predicted Pseudo


probability for the respective CMM Level companies for business user
factors.

Table 3.24 portrays about the area that has been covered by the
business users of various CMM level companies. The probability range
is between 0.495 to 0.569.

Fig 3.20 depicts about the importance and magnitude of each of


the business user factors and their criticality in the success of
Information Technology Service Projects. Table 3.25 lists about various
business user factors, most significant factors as per business users are
Edge on competitors and supports value creation and addition and the
least significant factor is helping in magnification oj'sales.

Management Personnel Perspective:

Fig 3.21 shows about the predicted Pseudo probability for the
respective CMM Level companies for management personnel factors.

Table 3.26 portrays about the area that has been covered by the
management personnel of various CMM level companies. The
probability range is between 0.519 to 0.557.
Fig 3.22 depicts about the importance and magnitude of each of
the management personnel factor and their criticality in the success of
Information Technology Service Projects. Table 3.27 provides the
detailed factors list. Most significant factor is work-life balance and the
least significant one is line of control.

Technical Personnel Perspective:

Fig 3.23 it can be concluded that the predicted Pseudo probability


for the respective CMM Level companies for technical personnel
factors.

Table 3.25 portrays about the area that has been covered by the
management personnel of various CMM level companies. The
probability range is between 0.516 to 0.581.

Fig 3.24 depicts about the importance and magnitude of each of


the technical personnel factor and their criticality in the success of
Information Technology Service Projects. Table 3.26 shows the detailed
factor list. Most significant factor is performance standards and the least
significant factor is employee motivation.

Table 3.24 Are under the Curve for Business users


Table 3.25 Business User Factors

Table 3.26 Are under the Curve for Management Personnel

Area Under the Cuwe


Area
I ,557
2 ,519
CMM Level 3 ,520
4 ,535
5 ,551

Risk valuation

1
Projecl terns say in resourcinfi
Effective decision making 1
S5
S6 /
Effective line of control
Providing support to colleagues
-- ..-I
S7 1 Peer support across boundaries
S8 Management Vision
S9 Intranet as a communication agent
S 10 Strategic Change Management
S11 Partner's consultation for open innovation.
S 12 work-life balance
S 13 Readiness for regular change
S 14 Management Involvement in users requirements
S 15 Consultation of technical personnel
S 16 Management confidence on technical resources
S 17 well defined flexible execution process
172
Fig 3.19 Predicted Pseudo probability vs. CMM level chart for business
users
1
2
3
4
5

i 4 4 i i
CMM Lavel

Fig 3.20 Normalized Importance of various factors for business users


Normallzed Importance
0% 20% 406 60% 00% 100%
I I I I
Fig 3.21 Predicted Pseudo probability vs. CMM level chart for
Management Personnel

0 ' 4 i

CMM Level

Fig 3.22 Normalized Importance of various factors for Management


Personnel
Normallzed Importance
0% to% 40% 60% BOY 1 DO%
I I 1 I I

I I I I
0.00 0.02 Q 04 0 06 0 08
Importance
Fig 3.23 Predicted Pseudo probability vs. CMM level chart for Technical
Personnel
1
2
3
4
5

2 3 4 5

CMM Level

Table 3.28 Are under the Curve for Technical Personnel


Area Under the Cuwe

CMM Level
4 ,581
5 ,563
Table 3.29 Technical Personnel Factors

Fig 3.24 Normalized Importance of various factors for Technical


Personnel
Normalized Impottance
0% 20% 4MC 80% BOW 1WW
t I I I I I

I
I
o.oa 0.02
I I
0.04
1
0.06
1
0.06
1
0.10

Importance
Independent Variable Importance

The importance of an independent variable is a measure of how


much the network's model-predicted value changes for different values
of the independent variable. Normalized importance is simply the
importance values divided by the largest importance values and
expressed as percentages and can be seen in Table 3.27.

Table 3.30 Independent Variable Importance

Fig 3.25 Normalized Importance's


Normalized Importance

I t I
0.0 01 , 01 03
Importance
The importance chart is simply a bar chart Fig 3.25 of the values
in the importance table, sorted in descending value of importance. It
appears that variables related to a Projects success have the greatest
effect on how the network classifies success in Information Technology
Projects.

Thus Multilayer Perceptron procedure, constructed a network for


predicting the probability that a given lnformation Technology Project is
successful or not.

Fig 3.25 depicts that business users Weightage of success has a


significant importance when compared with technical personnel and
management personnel and the age is of least significance in finding the
success rate of Information Technology Enterprise Service projects.
References

1. Johansson, T. (2002). Strategies for continuous attention:


Newspaper Web Services and the creation of Mental Lock-ins,
Present at the British Academy of Management Conference, 9-1 1
September 2002, London.
2. Reengineering the new product development process: A
framework for innovation and flexibility in high technology firms
M.K. Malhotra, V. Grover, M. Desilvio.
3. Frederick Betz, 2011 - Managing Technological innovation
Competitive Advantage from change, 15-20.
4. David Aaker, 2006, Innovation: brand it or lose it
http://en.cnki.com.cn/Article~en/CJFDTOTAL-
GGDL20 1001003.htm
5. Albury, D. (2005). Fostering Innovation in Public Services.
Public Money & Management, 25 (1): 5 1-56.
6. Cooper, R.G. & Kleinschmidt, E.J. (1987) Success factors in
product innovation. Industrial Marketing Management, 16, 2 15-
223

Das könnte Ihnen auch gefallen