Volume 8, Issue 3, March 2023 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

A Hybrid Machine Learning Method for

Estimating Software Project Cost
Beatrice O. Akumba.,1 Iorshase Agaji.,2 Nachamada Blamah3 Emmanuel Ogalla2
1
Department of Mathematics/Computer Science, Benue State University Makurdi, Nigeria
2
Department of Mathematics/Statistics/Computer Science, Joseph Sarwuan Tarka University Makurdi, Nigeria
3
Department of Computer Science, University of Jos, Nigeria

Abstract:- This paper introduced the concept of a hybrid mathematics to carry out the estimation process. The
machine learning method for estimating software project software projects’ Source Lines Of Code (SLOC), function
cost. The literature review of some of the models points, the software cost drivers such as programming
commonly used in the software project cost estimation languages, risk assessments, design methodology and so on
was carried out. A hybrid method of algorithms are mostly used as inputs to the model. Some examples of
comprising Random Forest (RF), Kalman Filter (KF) algorithmic models include COCOMO, Functional Point
and Support Vector Machine (SVM) algorithms Analysis (FPA), and Putnam model among others. The non-
respectively were proposed to predict the software algorithmic models on the other hand which are known as
project cost and its completion time for software non- parametric models are estimation processes based on
projects. The proposed architecture of the model was deduction and analogy. They require the knowledge of
presented as well as the proposed the model. previously completed software projects that are similar to
the current software project for its estimations. Some
Keywords:- Software Cost Estimation, Machine Learning, examples include; expert judgment, Analogy based, Price to
Cost Estimation Models, Kalman Filter Algorithm, Support Win, Top-Down and Bottom-up models respectively. The
Vector Machine, Random Forest. learning-oriented models otherwise called the Machine
learning methods include the Artificial Neural Networks
I. INTRODUCTION (ANN), Fuzzy logic (FL), Analogy based, Bayesian
Software project cost estimation (SCE) entails Networks, Regression Tree, Genetic Algorithm (GA),
forecasting the entire cost required to undertake a software Support Vector Machines (SVM), Case-Based Reasoning
project system (Shukla, et al., [1]). The process is significant (CBR) and Kalman Filter Forecasting Methods (KFFM) [3].
because it is key to estimating the software project’s costs at The models have their merits and demerits and
the initial phase (Fadhil et al., [2]). Project cost estimations therefore are not the same with respect to their performance,
are important for computing resources and the budget such as inadequacy and inefficiency of the mathematical
required for a software project. The estimation of resources models to explain the nonlinear properties existing between
and schedule can be achieved through accurate estimation the input and output parameters of the software project cost
and forecasting. It is among the pivotal task in software estimation datasets. The shortcomings of algorithmic models
project development that entails forecasting the software led to the consideration of non-algorithmic models and
cost, software size and effort required to obtain a software intelligent techniques which are soft computing based
project system. (Chirra and Reza, [3]). (Resmi et al., [4]). All the aforementioned software cost
Software cost estimation requires the determination of estimation techniques/models are faced with the problem of
one or more of the following estimates; effort (usually in the absence of accuracy [2].
person-months), project duration (in calendar time) and cost Most of the existing software project cost estimation
(in money). Most of the software project cost estimation models are faced with challenges and there are two
models try to generate an effort estimate, which can then be substantial reasons behind the software project failure. They
converted into the project duration and cost. Although, effort are inappropriate project estimation in terms of size, costs
and cost are closely related, they are not necessarily related and personnel required and uncertainty about the software
by an easy changeover or conversion function. The software and system requirements due to advancement in technology.
effort is usually calculated as the person-months spent by With early and accurate estimate of project costs, client
the programmers, analysts and project managers on the expectations can be set and ensure that funds are made
actualization of the software project. The estimation of the available to achieve the project. There is need to have an
software effort can be represented in Naira or dollars by accurate estimate of the worth of a software project before
computing the average salary divided by the time spent by embarking on it as it is critical to the timely delivery and
the staff involved and further multiplied by the required subsequently return on investment. Therefore, this paper
estimated software effort. seeks to propose a hybrid machine learning method for
A lot of software cost estimation frameworks have estimating software project cost and to propose the use of
been established to enhance the estimation in the cost of non-algorithmic learning oriented technique of Kalman
software projects. The main categories of these models are Filter Forecasting Algorithm (KFA) to forecast the
algorithmic, non-algorithmic and learning oriented models. completion time of software projects.
The algorithmic models make use of equation and

IJISRT23MAR1673 www.ijisrt.com 2681

Volume 8, Issue 3, March 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
This paper is organized as follows: Section I is cost. However, it suggested that efforts should be made to
Introduction, Section. II outlines the objectives of the paper. use a combination of the estimation techniques (hybrid) in
Section III presents the literature reviewed. Section IV order to arrive at a better software project cost and quality
explains the Methodology while Section V is Conclusions estimate. Thus, to achieve reliable estimates, it is required to
and Future Work. have proper knowledge and understanding of each technique
and the relationship that exist between the software
II. OBJECTIVES attributes respectively. A combination of the methods would
greatly enhance accurate cost estimation of software projects
The specific objectives to be focused on this paper are: and would be seen to have addressed one of the complexes
 Carry out a literature review on machine learning and most challenging activity for the software companies
techniques of software project cost estimation Iqbal et al., [7] investigated the pros and cons of different
 Develop an architecture of the framework of the hybrid software cost estimation modeling techniques and concluded
software project cost estimation model. that one cost estimation method should not be stuck on. That
 Proposed a hybrid cost estimation model based on for cost and timeframe estimations, different types of
machine learning that will improve on the cost accuracy. estimation methods should be used (A Hybrid model),
comparison of the results and analysis of the variations
III. LITERATURE REVIEW should be explained, a documentation of the prediction
This section reviews relevant literatures with respect to made when making the cost estimates. Monitoring and
the concept of software project cost estimation; software evaluation of the entire project accuracy in software cost
project cost estimation methods/techniques and SCE based estimate process should be made. Also suggested was to
on machine learning methods. maintain all previous/past data of software projects. The
study concluded that it was very difficult to recommend a
A. THE CONCEPT OF SOFTWARE PROJECT COST fail-safe cost estimation method as most of the cost
ESTIMATION estimations techniques are scenario dependent. In order to
According to Fashina [5], software project cost select a method that suits any given project, a proper
estimation is a process that estimates/predicts the estimated analysis of the project is must be carried out. This is because
cost of the software project before the commencement of the a wrong estimation technique can significantly delay a
development. It explains the approximate requirements of software project’s delivery, and a right technique can make a
effort, development time and resources that are required to software project breeze through its deadlines. In addition, a
complete the software project. It is among the essential detailed analysis of some global factors must be made;
processes to start development for software by taking into otherwise, it becomes very difficult to give accurate
consideration all internal and external cost factors. The cost deadlines. Also, an emphasis on smaller details is necessary,
estimation is a tool that estimates the planning, budgeting as they can cause delays when they add up together.
and resource utilisation for the software projects. Before
cost estimation for a software project is done, the actual B. SOFTWARE COST ESTIMATION METHODS
requirements for the project, the complexity of the According to Iqbal et al., [7], the software project cost
requirements, and the software cost driver factors that affect estimation is broadly classified into three methods into
the development process (like, product factor, project factor, which these models are categorized and they are;
personal factor and hardware factor) must be considered. Algorithmic methods, Non-algorithmic methods and the
They act as the inputs to the cost estimation process. The Machine Learning methods.
output of the estimation process is Effort, Development Chirra and Reza [2] stated that the algorithmic method
Duration, and Resources. Figure 1 illustrates the software was developed to deliver a few equations mathematically
project cost estimation process diagrammatically showing its that can be used to carry out software cost estimation.
inputs and outputs.
The mathematical equations were based on historical
Effo data and research and made use of Source Lines of Code
SOFTWARE
Requirem
COST rt (SLOC), number of function calls to execute, cost drivers
ents ESTIMATION like programming language, design methodology, skill-
Development
PROCESS levels and risk assessment and so on as inputs to the model.
Time/Duration The algorithmic methods comprised the COCOMO,
Cost
factors Resour Function Point Analysis (FPA), Putnam model/SLIM
ces
respectively. The non-algorithmic models on the other hand
also known as non- parametric models make their estimation
Fig. 1: The Software Cost Estimation Process
process based on deduction and analogy. They require the
(Source: [5])
knowledge of previously completed software projects that
Shekhar et al., [6] explained the various models and are similar to the current software project for their
techniques used in determining the cost of the software. The estimation process. Examples include; expert judgment,
study tried to justify if there could be a best software cost Analogy based, Price to Win, Top-Down and Bottom-up
estimation method for software projects. An inference was models respectively.
made that there is as such not any single method that can be
regarded as the best method for estimating software project

IJISRT23MAR1673 www.ijisrt.com 2682

Volume 8, Issue 3, March 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
According to Aljohani and Qureshi, [8], the Non- compared with both the actual effort and the predicted effort
Algorithmic models are sub-classified into Learning based by applying Naïve Bayes, Logistic Regression and Random
models, Expert based Models and Dynamics Based models. Forest machine learning algorithms. The dataset used was
Examples of learning based methods are Artificial Neural surveyed by using the density function and projection plots
Networks, Genetic Algorithms, Fuzzy Logic, Bayesian to check both the nature and potential of the dataset after
Networks, Support Vector Regression, Regression Tree, which a number of models were built by applying the
Analogy and so on. The Expert based Model comprises Top- selected machine learning techniques. The built Naïve
Down estimation, Bottom-Up estimation, Expert Judgment Bayes, Logistic Regression and Random Forests models
and Price-to-Win estimation methods. The Dynamics based generated were further tested using five folds cross-
models are techniques that are used to approximate the validation and the models performance was evaluated with
changes in effort in the course of software development of a Area Under the Curve (AUC), CA, Precision and Recall,
system. The model is built such that it predicts changes in Confusion Matrix and ROC. The results of the investigation
the cost; the staffing team needs, schedules and the time of confirmed the validity of data mining as an alternative for
project as far as the main initial values of the project are COCOMO. Although, it was observed that no hybrid of the
made available to estimate. ML techniques were done and recommended that more ML
techniques should be adopted and used as well as extending
C. SCE BASED ON MACHINE LEARNING METHODS the research to cover more machine learning techniques and
Pospieszny et al., [9] predicted the continuous cost and a hybrid of them if possible.
duration target of variable of software projects using
machine learning algorithms. Three effective machine Khazaiepoor et al., [12] combined different Machine
learning predictive algorithms of Support Vector Machine Learning (ML) methods during several phases to provide a
(SVM), Multi-layer Perceptron Artificial Neural Network software development effort estimation model that
(MLP-ANN) and Generalised Linear Models (GLM) were irrespective of its high efficiency in estimating software was
used. This was achieved by using an International Software totally independent on the type of dataset used. Six datasets
Benchmarking Standards Group (ISBG) dataset after being were used to test the models independence in effort
preprocessed through smart data preparation. The input estimation. The developed model comprised three phases;
dataset was used for modeling by the application of three phase one combined the genetic algorithm and neural
machine learning algorithms (SVM, MLP, GLM) ensemble network to perform the feature selection operation. The
averaging and cross validation. The obtained model results second phase employed the genetic algorithm and imperial
for cost and duration estimation served as decision support competitive algorithm (ICA) to perform a clustering
system for organizations that are into software development operation. The third phase applied neural network to
and implementation systems. perform modeling and testing in clusters. Evaluation metrics
of MMRE, MDMRE and PRED(0.25) were applied to know
Resmi et al., [4] deployed an innovative process of the accuracy of the model and the learning rate was
Fuzzy analogy algorithm connected through the firefly compared with those of other models and the model was
algorithm to estimate the effort of software projects. The seen to have outperformed all the other methods for all the
input dataset used was from the Promise data repository datasets used.
comprising four datasets of NASA93, NASA60,
COCOMO81 and Deshnaris datasets respectively. A Abdulmajeed et al., [13] adopted machine learning
combination of the fuzzy analogy by the side of the firefly (ML) techniques in solving the problem of predicting the
and the Expectation-Maximisation (EM) process for the cost of the software project as it tries to give an estimate as
estimation of software project enhanced the accuracy level close as possible to the real cost of the project. The study
of the prediction. The fusion of fuzzy and firefly algorithm was based on the completed projects in the NASA (National
on clustered datasets improved the accuracy of the Aeronautics and Space Administration) data set, which
estimation to a considerable level. contained the attributes of 60 software projects. It also tried
to find and define a suitable method to give the closest
Arora and Mishra [10] proposed an estimation model possible estimate to the true cost. The K-Nearest Neighbors
for software cost evaluation using Artificial Neural Network KNN algorithm was used and applied to the NASA dataset,
(ANN) Multilayer Feed forward Neural Network using and its results were compared with both Cascade Neural
perception learning algorithm and COCOMO dataset. The Network and Elman Neural Network. The results
experiment was performed on 14 software projects with the performance was assessed and compared with the aid of
actual effort of the model compared with the experimental performance standards such as; MMRE, RMSE, and BRE.
effort values. The result indicated that accurate values were The results of the K-Nearest Neighbors (KNN) algorithm
attained by the model. indicated good accuracy levels in the course of forecasting
the required cost of developing software projects. The CNN
BaniMustafa [11] investigated the validity of machine algorithm and the ENN network gave less accuracy levels in
learning techniques to predict software effort estimation by software project cost estimation.
proposing an alternative method of COCOMO. This was
achieved by performing data mining on historical data to Zakaria et al., [14] used several algorithms of machine
obtain better software estimation. The dataset was learning to estimate the effort of software project
preprocessed in order to calculate the effort using the development. An application known as SOFREST estimator
COCOMO model equation. The estimated effort was then was invented to explain how the software estimation process

IJISRT23MAR1673 www.ijisrt.com 2683

Volume 8, Issue 3, March 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
works and to ascertain the inputs required to yield the Wani et al., [17] proposed an analogy centered model
result/output. The system required that the user inserted a that was based on differential evolution (DE) exploration
software project’s five inputs which are; the Lines of Code process for software cost estimation to help the industry
(LOC) number, Size of Database (DATA), needed Software practitioners. The model was evaluated on 676 software
Reliability (RELY), the Execution Time Constraint (TIME) projects from five different data sets (Kemerer, ISBSG10,
and Main Storage Constraint (STOR). The system’s output Miyazaki, Desharnais and China). MRE and MAE were
was the estimation of the effort that was required by the used to evaluate the model. The model results achieved were
software project in person-months. Four Machine Learning notably better as compared to other benchmarked analogy-
algorithms were used and they are; Random Forest (RF), based estimations. The result of the experiment indicated
Linear Regression (LR), Regression Tree (RT) and Support that the approach particularly DE was able to find the best
Vector Machine (SVM). Each model’s performance was possible estimators rapidly and out-performed the other
measured by the Mean Magnitude Relatives Error (MMRE). widely used standard methods like base-line Analogy Based
Estimation (ABE0) and automatically transformed linear
Karimi and Gandomani [15] modeled the effort model (ATLM).
estimation needed for software development by using a
hybrid of Fuzzy Neural Network (FNN) and Differential Bushra and Kadam [18] developed an improved
Evolution (DE) algorithm. The model obtained a better and technique for software cost estimations in agile software
accurate estimation of the software projects’ effort that was development environment based on Support Vector Machine
a better off as compared to previous methods. The model and Naïve Bayes algorithm. The system used COCOMO
was called ANFIS-DE and had two phases. The first phase (Constructive Cost Model) to achieve better software project
was the training phase and was applied to obtain the optimal cost estimation. The Profit or loss estimation forecast for
parameter for ANFIS using the DE algorithm and the second new software project was done with the help of some
phase known as the test phase used the obtained optimal historical data of a software company. Machine learning
parameters to test and assess the model. The DolBat techniques of Naïve Bayes and Support Vector Machine
algorithm had a better performance in all the assessed were used to perform the prediction using historical data.
criteria than the other algorithms in experimental functions. The SEERA dataset was used to evaluate the behavior of the
MMRE and PRED were used as performance parameters to model. The result indicated that the forecasted profit and
assess the model. The accuracy was seen to be better as loss prediction had an accuracy of 86.59% and 24.80%
compared to other optimization algorithms adopted from respectively. The accuracy of the software effort calculation
genetic, evolutionary and meta-heuristics respectively. was higher by 95.06% as compared to the SVM, 93.45%. It
recommended a further study to increase the accuracy of the
Dashti et al., [16] proposed a weight optimization future prediction of the forecasted profit and loss of the
technique based on the learnable evolution model (LEM) in system more than 90% by adopting the hybrid algorithm in
analogy-based estimation (ABE) for software projects. The the machine learning with the 97-98% overall effort
work employed three control parameters of similarity calculation accuracy.
function, K-nearest neighbor, and solution function
suggested by the LEM algorithm to find the optimal weights Zakaria et al., [14] employed Support Vector Machine,
which were parsed into the model. The amount was Linear Regression, Regression Tree and Random Forest
estimated as the proposed cost for the project. The proposed machine learning algorithms on COCOMO dataset to
cost of the project was compared to the real cost of the estimate software project cost. This was done by developing
project and its magnitude of relative error (MRE) metric was an application known as SOFREST Estimator. . Random
calculated. Euclidean and Manhattan distances were Forest (RF), Linear Regression (LR), Regression Tree (RT)
employed to obtain the similarity function and the most and Support Vector Machine (SVM) algorithms were
common solution functions of median and mean were used applied on the COCOMO NASA 1, COCOMO NASA 2,
to obtain the proposed estimation value. Three datasets were COCOMO81 datasets respectively. The results of the
used. These were Desharnais, Maxwell and International experiments showed that Support Vector Machine and
Software Benchmarking Standards Group (ISBSG). The Random Forest algorithms impressively gave consistent
datasets were used to evaluate the model and were tested in results with the COCOMO datasets regardless of the number
different states. The datasets were divided into training and of effort attribute used. The experiment revealed that there
testing datasets and used 10-fold cross-validation method to was good performance and increase in the accuracy of
train the model. Three evaluation criteria, of MMRE, PRED Support Vector Machine for estimation of software project
(0.25), and MdMRE were used to evaluate and compare the effort.
model against other evolutionary algorithms like genetic
algorithm, differential evolution, and particle swarm Nawaz et al., [19] proposed a model to predict the
optimization and more desirable results were obtained. The effort of software cost estimation using genetic
results revealed that employing the model indicated programming, neural networks and genetic algorithms with
considerable improvement in estimating software cost a distributed programming project dataset. Six types of
estimation. classifiers were used to test the 13 algorithms. The
classifiers used were Random Forest, REPTree, M5P,
ZeroR, Decision Table, Input Mapped Classifier, Additive
Regression, IBK, KStar, Gussian Processes, Linear
Regression, Multilayer Perceptron, and SMOreg. The result

IJISRT23MAR1673 www.ijisrt.com 2684

Volume 8, Issue 3, March 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
indicated that Random Forest achieved the best results in the evaluated were stacking using a generalized linear model,
first experiment using the first procedure dataset while the stacking using decision tree, stacking using a support vector
three models of REPTree, Additive Regression and Kstar machine, and stacking using random forest. The results
scored the best results using the second procedure dataset. proved that the stacking with random forest gave the best
ZeroR method had the worst results using the first dataset results as compared with the single model approaches using
while some other methods including Multilayer Perceptron, the machine or deep learning algorithms and other ensemble
IBk, and Linear Regression performed poorly on the second techniques.
dataset.
IV. METHODOLOGY
Varshini et al., [20] presented software effort
estimation process using ensemble techniques and machine A Hybrid Machine Learning Method for Estimating
and deep-learning algorithms to solve the single model Software Project Cost based on the commonly used machine
approaches and ensemble approaches that were considered learning algorithms is proposed in Figure 2. The proposed
for software cost estimations. This was achieved by the use model is divided into the Calendar Time Prediction Phase
of averaging, weighted averaging, bagging, boosting, and using Kalman Filter Algorithm (A) and the Financial Cost
stacking ensemble techniques to create multiple models Estimation Models using Random Forest and Support
termed base-level classifiers which were combined to Vector Machine learning techniques (B). The
produce better predictions as compared to single-level implementation of this model is beyond the scope of this
models. The various stacking models considered and paper.
.

Selected Perform Data Preprocessing Split Data into Train

Datasets
& Test (70%: 30%)

Perform Feature Selection

A B
Build Kalman Filter Build Hybrid
Prediction Model Models of SVM
&RF

Predict Software Predict Software

Completion Time & Cost and Effort
Duration

Evaluate the Model

Prediction Results

Fig. 2: Proposed framework for Hybrid Software Cost Estimation

IJISRT23MAR1673 www.ijisrt.com 2685

Volume 8, Issue 3, March 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
V. CONCLUSION AND FUTURE WORK (IJSEA) 2017, Vol.8, No.6, DOI:
10.5121/ijsea.2017.8603.
This paper has made efforts to present an elaborate [9.] P. Pospieszny, B. Czarnacka-Chrobot, and A.
concept of software cost estimation. It reviewed software Kobyliński. An Effective Approach for Software
cost estimation techniques/methods and the different Project Effort and Duration Estimation with Machine
machine learning techniques used for predicting the Learning Algorithms, The Journal of Systems &
software development cost in order to increase an Software 2017. doi: 10.1016/j.jss.2017.11.066.
understanding of this area of research. The paper went on to [10.] S. Arora, and N. Mishra. Software Cost Estimation
propose a hybrid software cost prediction based on the Using Artificial Neural Network. Soft Computing:
commonly used machine learning algorithms. The proposed Theories and Applications, Advances in Intelligent
hybrid model is divided into the Calendar Time Prediction Systems and Computing 2018, 584, Springer Nature
Phase using Kalman Filter Algorithm (A) and the Financial Singapore Pte Ltd. M. Pant et al. (eds.),
Cost Estimation Models using Random Forest and Support https://doi.org/10.1007/978-981-10-5699-4_6
Vector Machine learning techniques (B). [11.] A. BaniMustafa. Predicting Software Effort
Estimation Using Machine Learning Techniques. 8th
For future work, the implementation of this proposed International Conference on Computer Science and
model will be conducted along with the comparison of Information Technology (CSIT) 2018. 978-1-5386-
proposed model with existing techniques with benchmarked 4152-1/18/$31.00 ©2018 IEEE . ISBN: 978-1-5386-
and new datasets. More advanced schemes such as ten-fold 4152-1 . American University of Madaba Kings
cross-validation and feature selection will be considered for Highway, Madaba, Jordan, +96253294444,
the estimation of the prediction accuracy. In addition, an [email protected]
evaluation of the proposed hybrid model and their [12.] M. Khazaiepoor, A. K. Bardsiri and F. Keynia. A
comparison with existing techniques will be conducted. Dataset-Independent Model for Estimating Software
REFERENCES Development Effort Using Soft Computing
Techniques. Sciendo Journal of Applied Computer
[1.] H. K. Shukla, S. V. Singh and R. B. Singh. Cost Systems. ISSN 2255-8691 (online) ISSN 2255-8683
Estimation of Software by ANFIS based Artificial (print) December 2019, vol. 24, no. 2, pp. 82–93
Intelligence Approach. International Journal of https://doi.org/10.2478/acss-2019-0011
Research and Development in Applied Science and https://content.sciendo.com
Engineering 2021. (IJRDASE) Available online at: [13.] A. A. Abdulmajeed, M. A. Al-jawaherry, and T. M.
www.ijrdase.com Tawfeeq. Predict the Required Cost to Develop
[2.] A. A., Fadhil, R. G. Alsarraj, and A. M. Altaie. Software Engineering Projects by Using Machine
Software Cost Estimation Based on Dolphin Learning. Ashraf Journal of Physics 2021:
Algorithm. IEEE Acess Volume 8, 2020. Digital Conference Series 1897 012029.
Object Identifier 10.1109/ACCESS.2020.2988867. [14.] N. A. Zakaria, A. R. Ismail, A.Y. Ali, M. Khalid and
[3.] S. M. R. Chirra, and H. Reza. A Survey on Software N. Z. Abidin. Software Project Estimation with
Cost Estimation Techniques. Journal of Software Machine Learning. International Journal of Advanced
Engineering and Application 2019, 12, 226-248. Computer Science and Applications, (IJACSA) 2021,
https://doi.org/10.4236/jsea.2019.126014. Vol. 12, No. 6.
[4.] V. Resmi, S. R. Vijayalakshmi and S. Chandrabose. [15.] A. Karimi, and T. J. Gandomani. Software
An effective software project effort estimation system Development Effort Estimation Modeling using a
using optimal firefly algorithm. Cluster Computing Combination of Fuzzy-Neural Network and
https://doi.org/10.1007/s10586-017-1388-0. © Differential Evolution Algorithm. International
Springer Science+Business Media, LLC, part of Journal of Electrical and Computer Engineering
Springer Nature 2017. (IJECE) Vol. 11, No. 1, February 2021, pp. 707~715
[5.] A. Fashina. Challenges in the process of Estimating ISSN: 2088-8708, DOI: 10.11591/ijece.v11i1.pp707-
Software Cost . Jyvaskyla: University of Jyvaskyla 715. http://ijece.iaescore.com
2021, 50 p. Information System Science, master’s [16.] M, Dashti, T. J. Gandomani, A. D. Hasanpoor, H.
thesis. Zulzalil, A. B. M. Sultan. LEMABE: a novel
[6.] S. Shekhar, and U. Kumar. Review of Various framework to improve analogy-based software cost
Software Cost Estimation Techniques. International estimation using learnable evolution model 2022.
Journal of Computer Applications 2016 (0975 – PeerJ Comput. Sci. 7:e800 DOI 10.7717/ peerj-cs.800
8887) Volume 141 – No.11. [17.] Z. H. Wani, J. I. Bhat, and K. J. Giri.. A Generic
[7.] S. Z. Iqbal, M. Idrees, A. B. Sana, and N. Khan. Analogy-Centered Software Cost Estimation Based
Comparative Analysis of Common Software Cost on Differential Evolution Exploration Process.
Estimation Modelling Techniques. Mathematical Computational Intelligence, Machine Learning and
Modelling and Applications. Vol. 2, No. 3, 2017, pp. Data Analytics 2021. The Computer Journal, Vol. 64
33-39. doi: 10.11648/j.mma.20170203.12 No. 3.
[8.] M. Aljohani and R. Qureshi. Comparative Study of [18.] Q. Bushra,. and A. Kadam. An Improved Technique
Software Estimation Techniques. International for Software Cost Estimations in Agile Software
Journal of Software Engineering & Applications Development using Soft Computing Techniques. IT

IJISRT23MAR1673 www.ijisrt.com 2686

Volume 8, Issue 3, March 2023 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
(Information Technology) in Industry 2021, Vol. 9,
No.2. ISSN (Online): 2203-1731
[19.] M. A. Nawaz, M. Kumar, M. D. Zaki, K. G. Sunil, R.
Prithvi, K. Neeraj, and B. D. Raja. A Methodology
for Software Cost Estimation Using Machine
Learning Techniques. International conference on
Recent Trends in Artificial Intelligence, IOT, Smart
Cities & Applications (ICAISC-2020) (May 27,
2020)
[20.] P. A. G. Varshini, A. K. Kumari, and V. Varadarajan.
Estimating Software Development Efforts Using a
Random Forest-Based Stacked Ensemble Approach.
Electronics 2021, 10, 1195. https://doi.org/10.3390/
electronics10101195.

IJISRT23MAR1673 www.ijisrt.com 2687