Abstraction
Software development attempt appraisal is the procedure of foretelling the realistic attempt required to develop and keep undertakings based on uncomplete and unsure input. Effort appraisal techniques are proved that they are undependable, inaccurate and unpredictable. Accurate and dependable attempt appraisal is a disputing undertaking in package technology. Effort estimations may be used as input to project direction programs, undertaking budget estimations, better resource planning and programming, investing analysis, pricing and command procedures. The of import point in attempt appraisal is to foretell the dependable resources for undertaking demands. The two major jobs in package development attempt appraisals are 1 ) High attempt appraisal 2 ) Low attempt appraisal. The low attempt appraisal may take to jobs such as hapless undertaking direction, bringings postponement, budget overproductions and hapless package quality. The excessively high attempt appraisal may take to lost concern chances and neglecting to do efficient usage of resources. However, still the attempt appraisal is non matured or standardized for dependable appraisal.
This research work focuses on the reusability in package development attempt appraisal. Generally, package reuse drama major function in developing quality package and increase productiveness. Lots of tools and engineerings support package reusability in the recent old ages. Software reuse saves the package attempt and improves productiveness. This work proposed a new attempt appraisal theoretical account called COREANN ( COCOMO Reusability in Effort appraisal based on ANN theoretical account ) . COREANN modify COCOMO II Reusability metric and utilize Artificial Neural Network ( ANN ) with enhanced RPROP for accurate attempt appraisal and concluding attempt is optimized with the heuristic optimisation algorithm such as Fake Annealing ( SA ) algorithm.
As the package development procedure supports more reusability in footings of codification, analysis, design, proving and certification, but these are non reflected in most of the attempt appraisal theoretical accounts. This leads to inaccuracies in attempt appraisal. COCOMO II is one of the largely used package attempt appraisal theoretical account in package industry. The reusability in COCOMO II outputs better consequences but it is undependable and inaccurate. Another trouble in COCOMO II is to happen the values of size parametric quantities. To get the better of these jobs, this research focused on two constituents of COCOMO II theoretical account. First, alternatively of utilizing RUSE cost driver, three new reuse cost drivers are introduced such as Reuse Veryhigh Level Language ( RVLL ) , Required Integrator for Product Reuse ( RIPR ) , Reuse Application Generator ( RAPG ) . Second, three cost drivers such as PEXE, AEXE, LTEX are combined into individual cost driver Personnel Experience ( PLEX ) in order to cut down the undertaking cost. This modified COCOMO II attempt appraisal is accurate and dependable. Finally, this truth degree is more improved with the aid of ANN and optimisation techniques.
ANN is the machine acquisition technique which is used to better the public presentation of web that fits on COCOMO II. Normally, ANN Resilient BackPropagation ( RPROP ) is good suited for all categorizes of undertakings. So the proposed theoretical account is implemented utilizing ANN with Enhanced RPROP algorithm. The proposed theoretical account provides better truth for attempt appraisal and reduces error degree. The proposed theoretical account is trained with one of the heuristic optimisation algorithm such as Fake Annealing algorithm. Finally, the public presentation is evaluated utilizing the proposed theoretical account for the set of package undertakings. The rating standards are applied for the comparing of proposed theoretical account public presentation with the bing COCOMO II theoretical account. Evaluation standards are Magnitude of Relative Error ( MRE ) , Mean Magnitude of Relative Error ( MMRE ) and PRED ( P ) for rating of package cost appraisal. The consequences show that the usage of proposed COREANN theoretical account attempt appraisal is dependable, truth and predictable.
List OF ABBREVIATIONS
S.NO
Term
Abbreviation
1
Aa
Appraisal and Assimilation
2
AAF
Adaption Adjustment Factor
3
AAM
Adaption Adjustment Modifier
4
ACAP
Analyst Capability
5
AEXE
Application Experience
6
Americium
Algorithmic Models
7
ANN
Artificial Neural Network
8
Astatine
Percentage of codification Automatically Generated
9
BP
Back Propogation
10
CBD
Component-Based Development
11
Fingerstall
Commercial-Off-The-Shelf
12
Centimeter
Code Modification
13
COCOMO
Constructive COst MOdel
14
COREANN
COcomo Reusability in Effort appraisal based on ANN theoretical account
15
CPLX
Merchandise Complexity
16
Diabetes mellitus
Design Modification
17
EAF
Effort Adjustment Factor
18
EJ
Expert Judgment
19
Em
Attempt Multipliers
20
Flex
Development Flexibility
21
GQM
Goal/Question/Metric
22
IM
Integration Alteration
23
KSLOC
Thousands of Source Lines of Code
24
LTEX
Language and Tool Experience
25
Milliliter
Machine Language
26
MMRE
Average Magnitude of Relative Error
27
MRE
Magnitude of Relative Error
28
NOP
New Object Points
29
PCAP
Programmer Capability
30
PCON
Forces Continuity
31
PEXE
Platform Experience
32
PLEX
Forces Experience
33
Autopsy
Person Months
34
PMAT
Process Maturity
35
PREC
Precedentedness
36
PRED
Percentage of Prediction
37
Goad
Productiveness Rate
38
PVOL
Platform Volatility
39
RAPG
Reuse Application Generator
40
RCPX
Product Reliability and Complexity
41
Trust
Required Software Reliability
42
RESL
Architecture/Risk Resolution
43
REVL
Requirements volatility
44
RIPR
Requeried Integrator for Product Reuse
45
RPROP
Resilient Back Propagation
46
Ruse
Required Reuse
47
RVLL
Reuse Veryhigh Level Languauge
48
SA
Fake Annealing
49
SCED
Agenda
50
SF
Scale Factors
51
SLIM
Software Lifecycle Model
52
STOR
Storage Constraint
53
SU
Software Understandability
54
UNFM
Unfamiliarity of coder with the codification
List OF TABLES
S.NO
Table Name
Page No
1
Table 1 – COREANN Cost Drivers
2
Table 2 – Comparison of Effort Estimation
3
Table 3 – Comparison of Effort Estimation Results in MRE
List OF FIGURES
S.NO
FIGURE NAME
Page No
1
Figure 1 – Basic theoretical account of ANN
2
Figure 2 – Comparison of Effort Estimation
3
Figure 3 – Comparison of Effort Estimation Results in MRE
Index
Chapter – I
Introduction
Aim
PROBLEM STATEMENT
Contribution
ORGANIZATION OF THESIS
CHAPTER – Two
LITERATURE REVIEW
2.1 EXPERT JUDGMENT
2.2 ALGORITHMIC MODELS
2.2.1 COCOMO MODEL
2.2.1.1 COCOMO II MODEL
2.2.2 SLIM MODEL
2.3 MACHINE LEARNING
2.3.1 ARTIFICIAL NEURAL NETWORK
2.3.1.1 BASIC MODEL OF ANN
2.4 OPTIMIZATION TECHNIQUE
CHAPTER – Three
3.1 PROPOSED MODEL – COREANN
3.1.1 INTRODUCTION OF NEW METRICS
3.1.2 METRICS DEFINITION METHODOLOGY
3.1.3 THEORETICAL AND EMPRICAL VALIDATIONS
3.1.4 IMPLEMENTING ANN MODEL
3.1.4.1 ENHANCED RPROP
3.1.5 OPTIMIZATION BY SIMULATED ANNEALING
3.1.6 IMPACT OF REUSABLITY ON COREANN EFFORT ESTIMATION
Chapter – Four
4.1 EXPERIMENTAL RESULTS
4.2 PERFORMANCE AND EVALUATION CRITERIA
Chapter – Volt
5.1 CONCLUSION AND FUTURE WORK
Chapter – I
1.1 Introduction
Software attempt appraisal is one of the most critical and complex undertaking in package technology, but it is an inevitable activity in the package development processes. Over the last three decennaries, a turning tendency has been observed in utilizing assortment of package attempt appraisal theoretical accounts in diversified package development processes. Along with this enormous growing, it is besides realized that the essentialness of all these theoretical accounts in gauging the package development costs and fixing the agendas more rapidly and easy in the awaited environments. Although a great sum of research clip and money have been invested to bettering truth of the assorted appraisal theoretical accounts. Due to the built-in uncertainness in package development undertakings such as complex and dynamic interaction factors, alteration of demands, intrinsic package complexness, force per unit area on standardisation and deficiency of package informations, it is unrealistic to anticipate really accurate attempt appraisal of package development processes [ 1 ] . Many package undertakings were failed because of transcending their original estimations. A Recent studies on attempt appraisal studies stated that 60-80 per centum of package undertakings encounter attempt overproductions. Effort overruns normally lead to be overproductions and lost agenda. This would do deficiency of productiveness or loss of concern [ 2 ] . The most important end in package development attempt appraisal is to precisely foretell the sum of needed resources before the beginning of the undertaking. Many appraisal theoretical accounts are used to turn out truth but none of them have able to turn out that it was accuracy in all instances of package applications. [ 3 ]
Software reuse has been given great importance in package development for decennaries. Software reuse is the systematic pattern of developing package from the similarities in demands, architecture and sphere between applications that can be exploited to accomplish significant benefits in productiveness, quality and concern public presentation. Software reuse has benefits such as decreased attempt, improved productiveness, decreased time-to-market and decreased cost. There are different signifiers of reclaimable assets: package design, constituent libraries, COTS ( Commercial-Off-The-Shelf ) constituents, faculties in a domain-specific model, package architectures and their constituents organizing a merchandise household, undertaking and trial paperss. Component-Based Development ( CBD ) provides techniques for the decomposition of a system into independent parts conforming to a constituent theoretical account, thereafter composing of systems from pre-built constituents. CBD advocates the acquisition and integrating of reclaimable constituents. Components are more farinaceous than objects, which may be an advantage in recovering and assembly, and they conform to a constituent theoretical account, which facilitates composing. Several tools and engineerings are developed for package reuse and CBD. The impact of these engineerings on package quality, agenda or cost is really high. This research work addresses the significance of reusability in attempt appraisal and formulates new prosodies for reusability to find the dependable and accurate attempt estimations.
1.2 Aim
The purpose of the proposed theoretical account is to better the truth and predictable of package attempt utilizing reusability in package development.
To analyse the truth and predictability of reusability in COCOMO II and look into the cogency of reuse cost drivers for better public presentation.
To develop new attempt appraisal theoretical account based on COCOMO II and ANN theoretical accounts for better truth.
To explicate and formalize the new theoretical account COREANN by modifying COCOMO II cost drivers and ANN with enhanced RPROP algorithm.
To optimise the estimated attempt utilizing Simulated Annealing algorithm.
To analyse the effectivity of the proposed theoretical account by formalizing the attempt with COCOMO II and Magnitude of Relative Error ( MRE ) , Mean Magnitude of Relative Error ( MMRE ) and PRED ( P ) for rating of package cost appraisal.
1.3 PROBLEM STATEMENT
Reusability in package development attempt appraisal is possible merely the COCOMO theoretical account attempt appraisal. Reusability in COCOMO II concedes the better executing but it besides unreliable. Finally, to acquire the attempt in COCOMO II are undependable and unpredictable. In COCOMO II based on ANN theoretical account is reduced an mistakes in attempt but it besides non render accurate attempt appraisal. So overcome these jobs to present some reuse cost drivers in COCOMO II based on ANN theoretical account with enhanced RPROP. And to gauge the optimum parametric quantities for package attempt utilizing fake tempering algorithm. This new proposed theoretical account is called a COREANN.
1.4 CONTRIBUTIONS
To get the better of these bing system jobs to follow two ways in proposed COCOMO theoretical account. First, alternatively of RUSE cost driver to present three new reuse cost drivers such as Reuse Veryhigh Level Language ( RVLL ) , Required Integrator for Product Reuse ( RIPR ) , Reuse Application Generator ( RAPG ) . These cost drivers are increased reusability in package attempt and to cut down the cost in package development. Using this proposed theoretical account COREANN, to measure the package attempt accurate and dependable. But to better this truth and to cut down an mistakes in attempt with the aid of Artificial Neural Networks ( ANN ) technique. Finally, to gauge the optimal solution of the proposed theoretical account utilizing Simulated Annealing algorithm.
1.5 ORGANIZATION OF THESIS
The remainder of the paper contains as follows: Chapter II presents the Literature Review of the research work. In this chapter a brief description of assorted methods for cost & A ; attempt appraisal classs and optimisation technique. Chapter III describes the architecture of the proposed theoretical account with enhanced RPROP and SA algorithm. Chapter IV presents the experimental consequences for COREANN and COCOMO II. Finally, Chapter V concludes the research work and yields the decision & A ; future work
CHAPTER – Two
LITERATURE REVIEW
2.1 EXPERT JUDGMENT ( EJ )
2.2 ALGORITHMIC MODELS ( AM )
2.3 MACHINE LEARNING ( ML )
2.4 OPTIMIZATION TECHNIQUE
At an early phase in a undertaking ‘s life rhythm permit the development organisations and undertaking directors are pull off the resources efficaciously. In last 30 old ages, assorted methods for cost and attempt appraisal have been proposed in three classs:
Adept Judgment ( EJ )
Algorithmic Models ( AM )
Machine Learning ( ML )
In earlier period, some research workers found that utilizing more than one technique can cut down hazard of dependability and better the truth and anticipations. Further, utilizing more than one method may avoid the loss of utile information that other methods can supply. [ 4 ] [ 5 ]
So, utilizing a combination of method seems to be a solution for supplying more trustable determinations in package attempt appraisal. Harmonizing to some study a combination of single methods has seldom been used to gauge package attempt. However it has been implemented successfully in other scientific Fieldss. [ 3 ] [ 6 ] [ 7 ]
2.1 EXPERT JUDGMENT ( EJ )
In algorithmic theoretical accounts, Expert judgement is an effectual estimating tool on its ain or as an adjusting factor. The agencies of deducing an estimation are non denotive by using EJ. Therefore hard to reiterate [ 4 ] .
2.2 ALGORITHMIC MODELS ( AM )
Algorithmic theoretical accounts are used to keep the relationship between attempt and one/more undertaking features. The chief “ cost driver ” used in such a theoretical account is normally some impression of size. AM need standardization or to be adjusted to local fortunes. Examples of AM are the COCOMO and the SLIM theoretical accounts [ 4 ] .
2.2.1 COCOMO MODEL
COCOMO was originally published in 1981 ( COCOMO 81 ) , and became one of most popular parametric cost appraisal theoretical accounts of the 1980s [ 8 ] . It is uniting two type of methods such as parametric appraisal equation and burdening method. Based on the estimated instructions, the attempt is calculated by the attempted quality and the productiveness factors. COCOMO is based on the traditional topdown scheduling and concentrate on the figure of instructions. It is constituted of three degrees [ 4 ] :
1 ) Basic COCOMO:
By agencies of parametric appraisal equations, the estimation attempt and continuance are calculated on the footing of the estimated DSI. The dislocation to stages is realized in per centums. In this connexion it is differentiated harmonizing to system types such as organic-batch, semidetached-on-line, embedded-real-time and undertaking sizes [ 4 ] . The basic COCOMO theoretical account computes attempt as map of plan size, and it is same as individual variable method [ 9 ] .
Effort =a*sizeb
Where a and B are the set of values depending on the complexness of package. For the organic type of undertakings a=2.4, b=1.05, semi-detached type of undertakings a=3.0, b=1.12, Embedded type of undertakings a=3.6, b=1.2 [ 9 ] .
2 ) Intermediate COCOMO:
The appraisal equations are now taking into consideration 15 influence factors ; these are merchandise properties, computing machine properties, forces properties and undertaking properties. The grade of influence is categorized as really low, low, normal, high, really high, excess high [ 4 ] . An intermediate COCOMO theoretical account attempt is calculated utilizing a map of plan size and set of cost drivers or attempt multipliers [ 9 ] .
Effort = ( a*sizeb ) *EAF
where a and B are the set of values depending on the complexness of package and EAF ( Effort Adjustment Factor ) which is calculated utilizing 15 cost drivers. Each cost driver is rated from ordinal graduated table runing from low to high. For the organic type of undertakings a=3.2, b=1.05, semi-detached type of undertakings a=3.0, b=1.12, Embedded type of undertakings a=2.8, b=1.2 [ 9 ] .
3 ) Detailed COCOMO:
In this elaborate COCOMO the dislocation to stages is non recognized in per centums but by agencies of influence factors allocated to the stages. Besides, it ‘s separated harmonizing to the three degrees of the merchandise hierarchy. product-related influence factors are now taken into the duplicate appraisal equations [ 4 ] . In elaborate COCOMO the attempt is calculated as map of plan size and a set of cost drivers given harmonizing to each stage of package life rhythm. The stages used in elaborate COCOMO are demands be aftering and merchandise design, elaborate design, codification and unit trial, and integrating proving [ 9 ] .
Effort = ( a*sizeb ) *EAF*sum ( Wi ) .
The life rhythm activities like demand planning, system design, elaborate design, codification and unit testing, integrating and proving. In all above three theoretical accounts the factors a and B are depend on the development manner [ 9 ] .
2.2.1.1 COCOMO II MODEL
COCOMO II was published ab initio in the annals of package technology in 1995 with three bomber theoretical accounts ; an application-composition theoretical account, an early design theoretical account and a post-architecture theoretical account. COCOMO II has, as an input, a set of 17 Effort Multipliers ( EM ) or cost drivers which are used to set the nominal attempt ( PM ) to reflect the package merchandise being developed [ 8 ] .
1 ) The Application Composition Model
It uses object points for sizing instead than the size of the codification. The initial size step is determined by numbering the figure of screens, studies and the 3rd coevals constituents that will be used in application [ 9 ] .
Effort = NOP/PROD
Where NOP ( New Object Points ) = ( object points ) * ( 100- % reuse ) /100, PROD ( Productivity Rate ) =NOP/PersonMonths [ 9 ] .
2 ) The Early Design Model
It uses to measure alternate package system architectures where unadjusted map point is used for sizing [ 9 ] .
Effort = a*KLOC*EAF
Where a is set to 2.45, EAF is calculated as in original COCOMO theoretical account utilizing seven cost drivers ( RCPX, RUSE, PDIF, PERS, PREX, FCIL, SCED ) . Ruse: Reuse is consider as one factor, but it is a major factor for attempt appraisal [ 9 ] .
3 ) The Post Architecture Model
It is used throughout the care and existent development of a merchandise. The station architecture theoretical account lets in a set of 5 scale factors and 17 cost drivers [ 9 ] .
Effort= ( a*sizeb ) *EAF
Where a=2.55 and B is calculated as b=1.01+0.01*SUM ( Wisconsin ) , wi= amount of leaden factors [ 9 ] .
2.2.2 SLIM MODEL
Larry Putnam of Quantitative Software Management developed The Software Lifecycle Model ( SLIM ) in 1970 ‘s. [ 9 ] SLIM is based on the construct of Norden-Rayleigh curve which represents work force as a map of clip. The package equation for SLIM is defined as follows: [ 9 ]
S= E * ( Effort ) 1/3 * td4/3
Where td is the package bringing clip, E is the environment factor that reflects the development capableness, which can be derived from historical informations utilizing the package equation [ 9 ] . The size S is in LOC and the Effort is in person-year. Another of import relation is [ 9 ]
Effort = D0 * td3
Where D0 is a parametric quantity called manpower build-up which ranges from 8 ( wholly new package with many interfaces ) to 27 ( rebuilt package ) [ 9 ] . Uniting the above equation with the package equation, we obtain the power map signifier: [ 9 ]
Effort = ( D04/7*E-9/7 ) *S9/7 and
td = ( D0 -1/7*E-3/7 ) *S3/7
SLIM is widely used in pattern for big undertakings ( more than 70 KDLOC ) and SLIM is a package tool based on this theoretical account for cost appraisal and work force programming [ 9 ] .
2.3 MACHINE LEARNING ( ML )
Machine larning techniques have been used as a replacement to EJ and AM. Examples include fuzzed logic theoretical accounts, arrested development trees, nervous webs, and instance based logical thinking. [ 4 ]
2.3.1 ARTIFICIAL NEURAL NETWORK
ANN ‘s are normally formed from many 100s or 1000s of simple treating units, connected in analogue and feeding frontward in several beds. In a biological nervous web, the memory is believed to be stored in the strength of interconnectednesss between the beds of nerve cells. Using nervous web nomenclature, the strength or influence of an interconnectedness is known as its weight. ANN borrows from this theory and utilizes variable interconnectednesss weights between beds of fake nerve cells [ 10 ] .
ANN ‘s were proposed early in 1960 ‘s, but they received small attending until mid 80 ‘s. Prior to that clip, it was non by and large possible to develop webs holding more than two beds. These early two beds webs were normally limited to showing additive relationships between binary input and end product characters. Unfortunately, the existent universe is linear and does n’t impart itself to a simple binary theoretical account. The existent interruption through in ANN research came with the find of the back extension method [ 10 ] .
Because of fast and cheap personal computing machines handiness, the involvement in ANN ‘s has blossomed. The basic motivation of the development of the nervous web was to do the computing machines to make the things, which a human being can non make. Therefore, ANN is an effort to imitate a human encephalon. Hence, the ANN architecture can be easy compared with the human encephalon [ 10 ] .
2.3.1.1 BASIC MODEL OF ANN
Serious attempts to make a theoretical account of a nerve cell have been underway during the last 100 old ages, and singular advancement has been made. Nervous connexions are significantly fewer and similar than the connexions in the encephalons. The basic theoretical account of ANN is every bit shown in Figure. This is the combination of perceptron as discussed earlier which forms the Artificial Neural Network which is really practiced [ 10 ] .
Figure 1 – Basic theoretical account of ANN
Input Layer Hidden Layer Output Layer
An thought consists in the usage of a theoretical account that maps COCOMO theoretical account to a nervous web with minimum figure of beds and nodes to increase the public presentation of the web. The nervous web that we have used to foretell the package development attempt is the individual bed provender frontward nervous web with the individuality map at both the input and end product units. Two different larning algorithms back extension and RPROP are used to develop the web to happen the best acquisition algorithm. BP is used merely for little size of undertakings. But RPROP is used for all sort of undertakings. To utilize COCOMO dataset to develop and to prove the web and it was observed that the nervous web theoretical account with RPROP provided significantly better cost appraisals than the appraisal done utilizing COCOMO theoretical account [ 11 ] [ 12 ] [ 13 ] [ 14 ] .
2.4 OPTIMIZATION TECHNIQUE
Fake tempering ( SA ) is a random-search technique for determine the optimal solution in package undertaking ; In 1983 SA was developed for covering with extremely nonlinear jobs. An of import feature of the SA algorithm is that it does non necessitate specializer cognition about how to work out a peculiar job. This makes the algorithm generic in the sense that it can be used in a assortment of optimisation jobs without change the basic construction of the calculation. There are two of import standards for SA: Choice of Neighbor and Cooling strategies. Accuracy and optimisation of SA depends on these two standards. In Neighboring Solution, aggregation of all the optimum and non-optimal solutions is called Solution Space. During tempering procedure, algorithm indiscriminately selects any solution from these Solution infinite. [ 15 ]
Evolutionary algorithms, simulated tempering and taboo hunt are widely used heuristic algorithms for combinative optimisation. The term evolutionary algorithm is used to mention to any probabilistic algorithm whose design is inspired by evolutionary mechanisms found in biological species [ 16 ] [ 17 ] .
One of the best heuristic algorithms is simulated tempering ( SA ) algorithm. SA feats an analogy between the way in which a metal cools and freezes into a minimal energy crystalline construction and the hunt for a lower limit in a more general system. In the optimisation procedure, the solution indiscriminately walks in its vicinity with a chance determined by Metropolis rule while the system temperature decreases easy ; when the tempering temperature is shuting nothing, the solution stays at the planetary best solution in a high chance [ 16 ] [ 18 ] .
In package undertakings, Simulated Annealing algorithm can be used to gauge the optimum parametric quantities of the attempt constituents. The upper and lower bounds of the hunt infinite should be considerately given by interior decorator or be cited from other mention documents, if possible. By and large talking, if a larger hunt infinite is built, it would be more clip of calculations and convergence of hunt may go really slow. Conversely, if the hunt infinite is set excessively little, the optimum parametric quantities likely could non been found [ 16 ] .
CHAPTER – Three
3.1 PROPOSED MODEL – COREANN
The purpose of this research is to develop the proposed COCOMO II based on ANN theoretical account utilizing reusability to foretell the attempt more accurately and dependability. In COCOMO II, to happen the values of size parametric quantities are hard. But proposed theoretical account COREANN is supplying the best consequence for gauging size utilizing alternatively of RUSE cost driver, to present three new reuse cost drivers..
The proposed theoretical account COREANN Effort Estimation expression is given below,
.
— — — — — — — — — 1
[ Non – Linear Equation ]
Where,
— — — — — — — — — 2
In proposed theoretical account COREANN, Scale Factors are same as the COCOMO II theoretical account graduated table factors such as PREC, FLEX, RESL, TEAM, PMAT.
— — — — — — — — — 3
— — — — — — — — — 4
— — — -5
Where,
PM = Person Month
EMi = Effort Multipliers ( Cost Drivers )
SFj = Scale Factors.
REVL = Requirements volatility
KSLOC= Thousands of Source Lines of Code
AT = Percentage of codification Automatically Generated.
AAM = Adaption Adjustment Modifier
AA = Assessment and Assimilation
AAF = Adaption Adjustment Factor
SU = Software Understandability
UNFM = Unfamiliarity of coder with the codification
IM = Integration Modification
Table 1 – COREANN Cost Drivers
Product dependability and complexness
RELY, DATA, CPLX, DOCU
Required reuse
RVLL, RIPR, RAPG
Platform trouble
TIME, STOR, PVOL
Forces capableness
ACAP, PCAP, PCON
Forces experience
PLEX
Facilities
TOOL, SITE
Required Development Schedule
SCED
To unite the three cost drivers such as PEXE, AEXE, LTEX into individual cost driver Personnel Experience ( PLEX ) for cut downing package cost.
3.1.1 INTRODUCTION OF NEW METRICS
Alternatively of RUSE metric, to present three new reuse prosodies are,
RVLL ( Reuse Very high Level Language )
RIPR ( Required Integrator for Product Reuse )
RAPG ( Reuse Application Generator )
3.1.2 METRICS DEFINITION METHODOLOGY
These new cost drivers are defined under GQM Methodology. GQM means Goal/Question/Metric. Goal is defined as every metric must hold a valuable end. Question is defined as every end must be answered by given inquiries and the inquiries are clearly addressed by the end. Metric is defined as the perfect reply for the given inquiries. It portion of the given end [ 19 ] .
GOAL OF THIS PROPOSED MODEL:
Object: Merchandise, Resource.
Purpose: Prediction, better
Focus: Attempt
Point of view: Undertaking Manager
Environment: 20 Undertakings Datasets
Question:
Q:1 ) How the integrating of merchandise would cut down both cost and attempt of the undertaking?
Q:2 ) Is the integrating giving accurate consequence?
Q:3 ) Is reusability needed?
Q:4 ) How the reusability in attempt appraisal addition truth and productiveness?
Q:5 ) Is reuse of high degree linguistic communication addition truth in attempt appraisal?
Q:6 ) How the staff accomplishments can be used to cut down cost?
Q:7 ) How mistakes impacting undertakings?
Q:8 ) What is the impact of interface in attempt appraisal truth?
Prosody:
For Q:1, Q:2.
The integrating of bing merchandise would better productiveness and give high quality of package merchandises.
For Q:3, Q:4, Q:5.
By the mention of 20 undertakings re-use of high-ranking linguistic communications cut down forces costs, cut down user defeats, and more rapidly fulfill user information demands within their sphere of pertinence. Overall these offer an highly attractive option for significantly bettering package productiveness.
For Q:6.
The overall package development accomplishments and abilities which our squad as a whole on mean brings to the merchandise integrating undertaking every bit good as experience with the specific application, platform, linguistic communication and tool.
For Q:7, Q:8.
Application generator addressed several system-oriented component-compatibility issues such as component interface conventions, informations structuring and plan controlled and error- handling conventions. So, Reusable of application generator can cut down mistakes and control plans in order to better the productiveness of package merchandises.
These replies are really utile to specify three new reuse prosodies decently.
3.1.3 THEORETICAL AND EMPIRICAL VALIDATIONS
These new cost drivers are decently validated with the aid of Theoretical ( Internal ) proof and Empirical ( External ) proof [ 20 ] .The of import of theoretical proof is to mensurate the buttockss of metric connotations utilizing with DISTANCE model and the empirical proof is garnering the information about the prosodies utilizing study method By utilizing 20 Projects datasets form company dataset, to formalize the EAF of proposed theoretical account. By seting the value of cost drivers, will give better consequence than past undertakings. Thus the proposed theoretical account validated by empirical proof. [ 21 ] [ 22 ] .
3.1.4 IMPLEMENTING ANN MODEL
COCOMO II attempt appraisal equation 1 should be transform non additive theoretical account to linear theoretical account by using natural logarithm on both sides, after that implementing to ANN theoretical account with Enhanced RPROP.
In ( PM ) = In ( A ) + 0.91 * In ( SIZE ) + SF1 * 0.01 * In ( SIZE ) + aˆ¦aˆ¦aˆ¦ . + SF5 * 0.01 * In ( SIZE ) + In ( EM1 ) + In ( EM2 ) + aˆ¦aˆ¦aˆ¦ + In ( EM17 ) — — — — — — — – 6
[ Linear Equation ]
OPest =WT0 + WT1 * IP1 + WT2 * IP2 + aˆ¦+ WT6 * IP6 + WT7 * IP7 +aˆ¦+ WT23 * IP23
— — — — — — — — — — — – 7
[ ANN Based Model For Effort Estimation ]
Where
OPest = In ( PM )
IP1 = 0.91 * In ( SIZE )
IP2 = SF1 * In ( SIZE ) , aˆ¦aˆ¦aˆ¦. , IP6 = SF5 * In ( SIZE )
IP7 = In ( EM1 ) , aˆ¦aˆ¦aˆ¦ , IP23 = In ( EM17 )
WT0 = In ( A )
WT1 = 1, aˆ¦aˆ¦aˆ¦aˆ¦. , WT23 = 1
IP1 to IP23 = & gt ; Inputs
OPest = & gt ; Output
WT0 = & gt ; Bias
WT1 aˆ¦aˆ¦ WT23 = & gt ; Weights ( Initial Value is 1 )
Actual ascertained attempt is compared with this estimated attempt. The difference between these values are the mistake in the attempt. It should be minimized.
3.1.4.1 ENHANCED RPROP ALGORITHM
In begininig of the algorithm, declared the initial values of the enhanced RPROP algorithm parametric quantities absolutely.
3.1.5 OPTIMIZATION BY SIMULATED ANNEALING
In proposed theoretical account COREANN, Simulated Annealing Algorithm is used to gauge the optimum parametric quantities of the package development attempt. The given solution method is helped to acquire optimum values of attempt:
Where,
EffortM = Measured Value of Effort,
EffortC = Computed Value of Effort harmonizing to the theoretical account used.
SIMULATED ANNEALING ALGORITHM
1. Let Ten: = initial config
2. Let E: = Eval ( X )
3. Let one = random move from the
moveset
4. Let Ei: = Eval ( move ( X, I ) )
5. If E & lt ; Ei so
Ten: = move ( X, I )
Tocopherol: = Ei
Else with some chance,
accept the move even though
things get worse:
Ten: = move ( X, I )
Tocopherol: = Ei
6. Goto 3 unless bored
The optimal solution for the proposed theoretical account utilizing Simulated Annealing Algorithm will be shown in future study.
3.1.6 IMPACT OF REUSABILITY ON COREANN EFFORT ESTIMATION
This research work focuses on the reusability in package development attempt appraisal. Normally, package reuse drama major function in developing quality package and increase productiveness. Software reuse saves the package attempt and improves productiveness. In this proposed COREANN theoretical account, to alter the COCOMO II reusability cost driver into three new reuse cost drivers and utilizing black box reuse methodological analysis. So no demand to alter the design and codification of bing theoretical account. This package reusability construct preserves the attempt appraisal and improves attempt truth degree. Finally, cost of the package development is besides decreasing with the aid of this reusability on COREANN
Chapter – Four
4.1 EXPERIMENTAL RESULTS
The proposed COREANN theoretical account is trained with ANN utilizing COCOMOII dataset for 20 undertakings and tested with the proof informations. The estimated consequence is comparing with COCOMO II.
Table 2 – Comparison of Effort Estimation
Undertaking
Unique Id
Actual Effort
Estimating Effort ( PM ) Using
COCOMO II
COREANN
1
205
117.6
212
2
211
117.6
189
3
40
31.2
34
4
24
36
27.3
5
43
25.2
49
6
15
8.4
5.2
7
9
10.8
15.2
8
36
25.2
43.9
9
277
352.8
261
10
95
72
108.5
11
67
72
121.3
12
39
24
27.4
13
255
360
276
14
77
36
85
15
288
215
292
16
345
360
322
17
398
360
417
18
299
324
319
19
102
60
92
20
76
48
63
Figure 2 – Attempt Estimation Comparison
-50
0
50
100
150
200
250
300
350
400
450
Undertaking Unique ID
Autopsy
Actual Effort
COCOMO II
COREANN
Actual Effort
205
211
40
24
43
15
9
36
277
95
67
39
255
77
288
345
398
299
102
76
COCOMO II
117.6
117.6
31.2
36
25.2
8.4
10.8
25.2
352.8
72
72
24
360
36
215
360
360
324
60
48
COREANN
212
189
34
27.3
49
5.2
15.2
43.9
261
108.5
121.3
27.4
276
85
292
322
417
319
92
63
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
4.2 PERFORMANCE AND EVALUATION CRITERIA
There are 20 undertakings will be used from company database for proving those dataset applying in proposed COREANN theoretical account. The undermentioned rating standard is used to measure and compare the public presentation of the proposed theoretical account with bing Cocomo II Model.
A common standard for the rating of cost appraisal theoretical account is the magnitude of comparative mistake ( MRE ) , and average magnitude of comparative mistake ( MMRE ) . MRE is defined as [ 11 ]
And Mean Magnitude of Relative Error ( MMRE ) for N undertakings is defined as [ 11 ]
Following to cipher the PRED ( P ) value. If lower MRE & A ; MMRE and higher PRED ( 25 ) , the package appraisal theoretical account attempt are more accuracy and predictable than other theoretical accounts [ 11 ] [ 23 ] [ 24 ] .
where K is the figure of undertakings where MRE is less than or equal to p ( usually p value is 25 % ) . [ 25 ]
Table 3 – Comparison of Effort Estimation Results In MRE
Undertaking
Unique Id
MRE Using
COCOMO II
COREANN
1
42.63
3.41
2
44.27
10.43
3
22
15
4
50
13.75
5
41.39
13.95
6
44
65.33
7
20
68.89
8
30
21.94
9
27.36
5.78
10
24.21
14.21
11
7.46
81.04
12
38.46
29.74
13
41.18
8.235
14
53.25
10.39
15
25.35
1.39
16
4.35
6.67
17
9.55
4.77
18
8.361
6.69
19
41.18
9.80
20
36.84
17.11
MMRECOCOMO = 30.592 PRED ( 25 ) COCOMO = 35.00
MMRECOREANN = 20.426 PRED ( 25 ) COREANN = 80.00
0
10
20
30
40
50
60
70
80
90
MRE
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
Undertaking Unique ID
Figure 3 – MRE Comparison
COCOMO II
COREANN
While comparing consequences of COCOMO II and COREANN, it is clearly shown that the proposed theoretical account COREANN works better than COCOMO II. That is, the COREANN theoretical account is estimated lower MRE & A ; MMRE and higher PRED ( 25 ) than the COCOMO II theoretical account. So the anticipation truth of COREANN is high.
Chapter – Volt
5.1 CONCLUSION AND FUTURE WORK
Now a yearss, about no theoretical account can gauge the package attempt with more truth and predictable. In this research work, proved the package attempt appraisal of COREANN are more accuracy, predictable and reliable.Because reusability methodological analysis is used in the COREANN theoretical account. That is, alternatively of RUSE costdriver, to present three new reuse cost drivers in proposed theoretical account. Normally, COREANN is based on ANN theoretical account with Simulated Annealing optimisation technique. So an mistakes of attempt appraisal are reduced utilizing ANN theoretical account. It ‘s besides utile for calculating attempt in all sort ( little and big size ) of undertakings. Finally, to find the optimal solution for proposed theoretical account utilizing fake tempering technique.
By utilizing rating standards, To compare the public presentation of COREANN with the COCOMO II attempt. In public presentation and rating standards subdivision, to demo the values of MRE, MMRE and PRED ( 25 ) for COREANN & A ; COCOMO II. Based on these calculation, clearly to turn out the public presentation of COREANN attempt appraisal is much better than COCOMO II. Because MRE & A ; MMRE values of COREANN is lower than COCOMO II and PRED ( 25 ) value of COREANN is higher than COCOMO II. In future, to compare the proposed theoretical account with all other lending attempt appraisal theoretical accounts and to turn out the proposed theoretical account appraisal will much better than all other appraisals.
