Optimization of software cost estimation model based on biogeography-based optimization algorithm

Abstract

Estimation of software cost (ESC) is considered a crucial task in the software management life cycle as well as time and quality. Prior to the development of a software project, precise estimations are required in the form of person month and time. In the last few decades, various parametric and non-algorithmic or non-parametric approaches regarding the estimation of software costs have been developed. Among them, the constrictive cost model (COCOMO-II) is a commonly used method for estimating software cost. To further improve the accuracy of this model, researchers and practitioners have applied numerous computational intelligence algorithms to optimize their parameters. However, accuracy is still a big problem in this model to be addressed. In this paper, we proposed a biogeography-based optimization (BBO) method to optimize the current coefficients of COCOMO-II for better estimation of software project cost or effort. The experiments are conducted on two standard data sets: NASA-93 and Turkish Industry software projects. The performance of the proposed algorithm called BBO-COCOMO-II is evaluated by using performance indicators including the manhattan distance (MD) and the mean magnitude of relative error (MMRE). Simulation results reveal that the proposed algorithm obtained high accuracy and significant error minimization compared to original COCOMO-II, particle swarm optimization, genetic algorithm, flower pollination algorithm, and other various baseline cost estimation models.

Keywords

Software cost estimation COCOMO-II biogeography-based optimization NASA-93 Turkish software project

1. Introduction

Estimating of software cost is one of the key element of the software management life cycle which is widely used for planning and managing project cost and time [1, 3]. Reliable estimation makes the software project more efficiently and accurately, such as resources allocation, reducing the project failures, level of reliability required, the tool used for a software project, the capability of the programmers, and so on [4, 5]. Several software cost estimation methods and procedures are proposed which are categorized into two sets: parametric or algorithmic and non-algorithmic or non-parametric. The most commonly used non-algorithmic methods are estimation via analogy, top-down, bottom-up estimation, Parkinsons law and price to win. Some of the prevalent parametric methods are COCOMO (Constructive Cost model), Baily Basil (BB), Walston Filex model (WF), Halstead, Doty, Putnam, function point analysis, SEL and COCOMO-II. COCOMO-II is the most widely used algorithmic estimation model which is discussed in Section 3. However, this model is still lacking in term of accuracy. To further improve the accuracy of this model, several algorithms have been studied. In this paper, the recently developed derivative free method, namely, the biogeography-based optimization algorithm (BBO) has employed for training the coefficients of COCOMO-II model and compared the proposed model with baseline model denoted by COCOMO-II. The other well-known derivative free methods including genetic algorithm (GA), and particle swarm optimization (PSO) are used to establish fair comparison of the proposed methodology and other various software effort estimation models likewise Doty, IVR, Halstead, Bailey Basil, and SEL. The two historical datasets, NASA-93 and Turkish Industry software projects were used in our carried out experiments.

The main contribution of this paper is given as:

•
BBO algorithm is employed for the first time for settling well optimized choice of parameters in respect of COCOMO-II for software cost estimation.
•
Experimental results of the proposed methodology are compared with several derivative free methods including GA, FPA and PSO versus other models of the same natures likewise Dosty, IVR, Halstead, Bailey Basil, and SEL.
•
Two different standard dataset NASA and Turkish Industry software projects are used to prove the effectiveness of the proposed methodology

The rest of this paper is arranged as follow: Section 1 presents the background of the study, in Section 2, we discuss some related work regrading estimation of software project, while Section 3 illuminates COCOMO-II model. The proposed methodology is discussed in Section 4. Furthermore, Sections 5 and 6 explain results analysis, conclusion and future recommendation.
2. Related work

Many estimation techniques were used to optimize the current coefficient of the COCOMO model [6, 7], such as PSO, GA, tabu search, intelligent water drop algorithms and many more. In [8], a cuckoo search algorithm was applied to optimize the coefficients of COCOMO-II. Experiments were conducted using 18 datasets from NASA-93 software projects. Experimental results demonstrate that the cuckoo search algorithm performs better than Baily-basil, Doty, Hallstead, and original COCOMO-II. Sweta and Pushkar [9] proposed a hybrid approach of cuckoo search and artificial neural networks to optimize the existing COCOMO-II parameters. They used NASA-93 projects for experimental work. A hybrid technique was presented in [10] for optimizing the parameters of COCOMO-II. The authors used tabu search with GA for the parameters optimization of COCOMO-II. Urbanek et al. [11] used analytical programming with differential evolution techniques for software cost estimation. Their results demonstrate that the proposed method outperforms differential evolution technique.

Dalal et al. [12] presented a generalized reduced gradient nonlinear optimization algorithm with best-fit analysis to improve the current coefficients of COCOMO-II. They applied the optimized COCOMO-II on NASA-93 software projects. Their results demonstrate that the optimized COCOMO-II gave better performance than COCOMO-II. Pereira et al. [13] proposed COCOMO-II for effort estimation using organization case study (aeronautical industry). They concluded that the COCOMO-II has better estimation technique for real software project as compared to other estimation techniques. Shepperd et al. [14] proposed a new method named as expert judgment to calculate the cost of software projects. Amelia Effendi et al. [17] proposed a bat algorithm for optimization of COCOMO-II. Sumera et al. [18] proposed a hybrid whale-crow optimized-based optimal regression method for estimation software project cost. They used four software industries dataset in their experiments. According to their experimental results, the proposed model has better than other some estimation model. Sheta et al. [19] proposed soft-computing method for estimation of software project cost. They used PSO to optimize the coefficients of COCOMO. They also used fuzzy logic to create a set of linear methods over the domain of possible software LOC. The proposed algorithms are compared with some baseline methods like, HS, WF, BB and Doty models. However, COMOCO-II is still requiring lacking of accuracy.

In this study, we offered a BBO for optimization of COCOMO-II coefficients. The proposed algorithm is compared with conventional COCOMO-II and some other baseline methods for NASA-93 and Turkish Industry software projects. Simulation results reveal that the proposed algorithm gives better results as compared to original COCOMO-II, PSO, GA and some other baseline models in terms of MMRE and MD.

Figure 1.

The process of COCOMO-II model.

3. COCOMO model

COCOMO model was first developed by Boehm in 1981 [20]. It is an algorithmic technique used for project cost or effort estimation. COCOMO models consists of three-layer, basic, intermediate, and detail layer. COCOMO-II is a new form of COCOMO and is proposed by B. B in 1995 [21]. It provides better cost estimation of the software projects compared to COCOMO. Figure 1 presents the overall process of COCOMO-II. The formula of effort estimation in terms of man month for projects is given as:

$\displaystyle MM=A\cdot\textit{Size}^{E}\cdot\prod_{i=1}^{17}\textit{EMi}$ (1)

Here $A$ is constant and its value is 2.94. Size is represented in a thousand source LOC (KSLOC). There are seventeen cost drivers known as effort multiplier (EM) which are given in Table 1.

Table 1

COCOMO-II model effort multipliers [21]

Drivers	Very low	Low	Nominal	High	Very high	Extra high
RELY	0.82	0.92	1.00	1.10	1.26	–
DATA	–	0.90	1.00	1.14	1.28	–
CPLX	0.73	0.87	1.00	1.17	1.34	1.74
RUSE	–	0.95	1.00	1.07	1.15	1.24
DOCU	0.81	0.91	1.00	1.11	1.23	–
TIME	–	–	1.00	1.11	1.29	1.63
STOR	–	–	1.00	1.05	1.17	1.46
PVOL	–	0.87	1.00	1.15	1.30	–
ACAP	1.42	1.19	1.00	1.85	0.71	–
PCAP	1.34	1.15	1.00	0.88	0.76	–
PCON	1.29	1.12	1.00	0.88	0.81	–
APLEX	1.22	1.10	1.00	0.90	0.81	–
APLEX	1.19	1.09	1.00	0.91	0.85	–
LTEX	1.20	1.09	1.00	0.91	0.84	–
TOOL	1.17	1.09	1.00	0.90	0.78	–
SITE	1.22	1.09	1.00	0.93	0.86	0.80
SCED	1.43	1.14	1.00	1.00	1.00	–

The exponent $E$ in Eq. (1) can be calculated as:

$\displaystyle E=B+0.01\cdot\sum_{j=1}^{5}\textit{SFj}$ (2)

Here $B=$ 0.91 is a constant term. The scale factors (SF) rating range from very low to extra high, which are given in Table 2.

Table 2

COCOMO-II model scales factors [21]

Scale factors	Very low	Low	Nominal	High	Very high
PREC	6.20	4.96	3.72	2.48	1.24
FLEX	5.07	4.05	3.04	2.03	1.01
RESL	7.07	5.65	4.24	2.83	1.41
TEAM	5.48	4.38	3.29	2.19	1.10
PMAT	7.80	6.24	4.68	3.12	1.56

4. Proposed methodology

There are numerous uncertainties in the software project cost estimation using COCOMO-II. In COCOMO-II the multiplicative constants A and B need be to optimized to get better estimation. The aim of this study is to optimize these constants to improve the performance of COCOMO-II by using BBO optimization method. The performance of the BBO-COCOMO-II is evaluated by comparing with PSO, GA, and other various cost estimation models, such as IVR, SEL, Bailey-Basil, Doty, and Halstead [22, 23, 24, 25]. NASA-93 and Turkish Industry software projects are used for the experiments. The input is project size in term of KLOC, measured effort, 17 effort multipliers, and 5 scale factors, and the output is in the form of new optimized values of A and B.

4.1 Fitness function

In terms of cost estimation for different projects, if the predicted effort approximately matches the real effort then projects are successfully completed. It means for higher accuracy, it’s needed the lower value of MMRE and MD. Therefore, our goal is to minimize MD and MMRE between the actual and predicted efforts. Fitness functions used in our experiments are given in Eqs (5)–(7).

4.2 BBO algorithm

BBO is a stochastic based computational algorithm, which was proposed by Simon in 2008 [25]. In BBO, habitants are considered as individuals which are generated randomly. Each individual has habitat suitability index (HSI) i.e. fitness value which shows its degree of goodness. Habitat with high HSI represents a good solution while habitat with low HSI presents a poor solution. Factors that can affect HIS of a habitat such as rainfall, diversity of vegetation, topographic diversity, land area, and temperature etc., are called suitability index variables (SIVs). Features emigrate from high HSI habitat to low HSI habitat. Two migration opera- tors named as emigration and immigration are used to evaluate and improve solutions of the given optimization problem. The emigration rate in high HSI habitat is high and its immigration rate is low, while in low HSI habitat, the immigration rate is high and the emigration rate is low [28, 29]. In BBO, first a population of habitats is randomly generated. Each habitat is denoted by an N-dimensional integer vector $[\textit{SIV}1\ldots\textit{SIVN}]$ SIVs (solution feature) represent genes while habitats represent chromosome. The HSI value of habitat is the fitness value function associated with that solution. The HSI value is calculated as

$\displaystyle\textit{HSI}=f(\textit{Habitat})=f[\textit{SIV}1\ldots\textit{% SIVN}_{\textit{var}}]$ (3)

Two operators’ migration (i.e., immigration and emigration) and mutation are used to get new population for next generation. Mutation factor is employed to the habitat which tends to increase the biological diversity of the population probabilistically. The mutation rate ‘m’ is calculated by the following formula.

$\displaystyle m=m_{\textit{max}}(1-p/p_{\textit{max}})$ (4)

Where “ $m_{\textit{max}}$ ” represents user-defined parameter. The Algorithm 1 illustrate the pseudo code of the BBO-COCOMO-II.

4.3 Evaluation criteria and data set

4.3.1 Data set description

Two datasets NASA-93 and Turkish Industry software projects [26, 27] were used for experiments. These software projects contain KLOC, actual effort, estimated effort, 17 cost drivers or effort multipliers (EM), and 5 Scale Factors (SF) with rating level from very low to extra high.

Table 3
Parameter settings in BBO algorithm

Parameters	Values
Maximum no. of iteration	1000
Habitats number	50
Keep rate	0.2
Alpha	0.9
mutation	0.1

Table 4

Parameter settings in genetic algorithm framework

Parameters	Values
Maximum no. of iteration	500
The size of population	50
Selection	Tournament
Crossover function	0.9
Mutation probability	0.015

Table 5

Parameter settings in PSO framework

Parameters	Values
Maximum no. of iteration	100
Size of population	50
Acceleration coefficients	[2.1, 2.]
Inertia weight	{1, 0.99]
Max. velocity	100

Table 6

Parameter settings in FPA framework

Parameters	Values
Maximum no. of iteration	100
Size of population	50
No. of independent run	50
Probability switch	0.8

Table 7

Magnitude of relative errors for estimations using BBO, COCOMO-II and others models using NASA-93 datasets

Project id	BBO		COCOMOII		PSO		GA		FPA		Doty		IVR		Halstead		B-Basil		SEL
1	0.	1409	0.	2643	0.	2204	0.	3200	0.	211	0.	2200	0.	3571	0.	2154	0.	7545	1.	038
3	0.	0098	0.	2008	0.	0211	0.	0766	0.	0666	0.	1396	0.	4365	0.	5206	0.	7005	0.	8816
6	0.	0817	0.	1786	0.	1684	0.	1949	0.	1929	0.	0024	0.	4373	0.	7281	0.	653	0.	6442
7	0.	3241	0.	0285	0.	3841	0.	0037	0.	0036	0.	3297	0.	8176	0.	5756	0.	5844	1.	178
10	0.	5234	0.	5969	0.	5595	0.	6093	0.	6733	0.	4803	0.	6909	0.	1304	0.	6847	1.	4673
16	0.	0084	0.	0774	0.	1680	0.	0350	0.	0344	0.	8844	0.	8238	0.	9444	0.	7184	2.	1919
19	0.	3778	0.	4811	0.	4131	0.	0046	0.	0033	0.	5949	0.	8768	0.	1528	0.	638	1.	6508
22	0.	0106	0.	1127	0.	1600	0.	3108	0.	2111	0.	4169	0.	43	0.	2682	0.	7683	1.	3904
23	0.	0457	0.	1775	0.	1419	0.	2579	0.	2655	0.	3886	0.	5241	0.	0653	0.	7284	1.	3236
27	0.	1446	0.	2858	0.	1947	0.	2736	0.	273	0.	1272	0.	3229	0.	396	0.	7457	0.	874
31	0.	0001	0.	1338	0.	1069	0.	1326	0.	1233	0.	0996	0.	1945	0.	2336	0.	7896	0.	8418
34	0.	1713	0.	3212	0.	1979	0.	2579	0.	111	0.	0048	0.	2279	0.	5305	0.	7528	0.	6639
35	0.	0767	0.	2327	0.	1249	0.	2059	0.	2055	0.	1786	0.	3968	0.	3889	0.	7284	0.	9576
37	0.	0049	0.	1678	0.	0425	0.	1269	0.	122	0.	0823	0.	2925	0.	4531	0.	747	0.	7963
39	0.	1170	0.	2861	0.	1297	0.	1811	0.	2	0.	1153	0.	111	0.	6229	0.	7694	0.	4612
40	0.	2818	0.	3993	0.	3257	0.	333	0.	227	0.	2767	0.	1545	0.	607	0.	8382	0.	2029
48	0.	3644	0.	4393	0.	4474	0.	5366	0.	454	0.	1436	0.	1951	0.	0906	0.	7986	0.	9228
49	0.	0420	0.	1879	0.	1176	0.	1299	0.	112	0.	0535	0.	1975	0.	3704	0.	7779	0.	7569
56	0.	1915	0.	2303	0.	3705	0.	5229	0.	4543	0.	44745	0.	3116	1.	0272	0.	816	1.	5183
57	0.	1448	0.	2312	0.	2765	0.	4083	0.	333	0.	6191	0.	6258	0.	4747	0.	7382	1.	733
58	0.	3281	0.	5116	0.	4269	0.	1624	0.	1442	0.	6414	0.	4362	0.	9288	0.	849	0.	206
63	0.	0131	1.	0779	0.	1579	0.	3223	0.	363	2.	7097	2.	6297	2.	6893	0.	4324	5.	2772
74	0.	8456	0.	8694	0.	8973	0.	8735	0.	7656	0.	5559	0.	4927	0.	7391	0.	7094	0.	2598
82	0.	2164	0.	3051	0.	3239	0.	4370	0.	434	0.	3189	0.	2963	0.	4373	0.	8828	0.	1463
84	0.	7666	0.	8009	0.	8971	0.	8004	0.	9	0.	6940	0.	6573	0.	8086	0.	9617	0.	4896
86	0.	6655	0.	7004	0.	7156	0.	7662	0.	674	0.	7667	0.	764	0.	793	0.	9558	0.	6067
87	0.	6128	0.	6519	0.	9724	0.	7321	0.	563	0.	7265	0.	7253	0.	7509	0.	972	0.	5384
88	0.	7616	0.	7892	0.	7934	0.	8274	0.	7873	0.	8411	0.	8349	0.	8714	0.	929	0.	7328
90	0.	8002	0.	8096	0.	8448	0.	8825	0.	877	0.	7814	0.	8061	0.	6968	0.	9374	0.	6266
93	0.	0424	0.	2628	0.	1109	0.	1144	0.	105	0.	6835	0.	5604	0.	0995	0.	8977	0.	4823
94	0.	61283	0.	7892	0.	7935	0.	8831	0.	2451	0.	7768	0.	8734	0.	1343	0.	8956	0.	5611
97	0.	761631	0.	8999	0.	8765	0.	8825	0.	8761	0.	9641	0.	5643	0.	8762	0.	7875	0.	8932
98	0.	875698	0.	8096	0.	8447	0.	6573	0.	1231	0.	23453	1.	5641	2.	3431	3.	6521	2.	761
99	0.	800297	0.	6595	0.	6187	4.	984	2.	41232	6.	9918	0.	7611	0.	8751	0.	8734	0.	8621
101	0.	593385	3.	978	5.	2567	0.	0145	0.	0126	0.	01235	0.	7642	0.	5142	0.	673	3.	451
MMRE	11.	76%	18.	9%	13.	844%	18.	262%	13.	56%	23.	30%	21.	15%	22.	35%	30.	14%	39.	389%

4.3.2 Evaluation criteria

To evaluate the accuracy of BBO-COCOMO-II and to compare it with other approaches including original COCOMO-II, we have used the most common evaluation methods MRE, MMRE and MD. MRE calculates the absolute percentage of errors among actual and expected effort for each reference software project. The parameter setting of the proposed BBO algorithm, GA, FPA, and PSO are given in Tables 3–6 respectively. The Matlab implementation of BBO introduced by Dan Simon [24] is applied in our paper.

$\displaystyle\textit{MRE}_{i}=\left|{\frac{\textit{Actual Effort}-\textit{% Estimated Effort}}{\textit{Actual Effort}}}\right|$ (5)

MMRE of software projects is given as

$\displaystyle\textit{MMRE}=\frac{1}{N}\sum_{i=1}^{n}\textit{MRE}_{i}$ (6)

MD is the difference between predicted effort and actual effort which can be calculated as

$\displaystyle MD=\left({\sum_{i=1}^{n}\textit{Actual Effort}-\textit{Estimated% Effort}}\right)$ (7)

Where $n$ represents the total software projects numbers.

Table 8

Magnitude of relative errors for estimations using BBO, COCOMO-II and others models using Turkish Industry datasets

Project id	BBO		COCOMOII		PSO		GA		FPA		SEL		Halstead		IVR		Doty		B-Basil
1	0.	2817	1.	9070	0.	2819	0.	3310	0.	1918	2.	2409	2.	0610	9.	0228	12.	9206	15.	3924
2	0.	0198	0.	4282	0.	0253	0.	3269	0.	0314	0.	3337	0.	0100	2.	7725	4.	4631	5.	1451
3	0.	3739	1.	0676	0.	3739	0.	7429	0.	3951	0.	1949	0.	3629	2.	9895	4.	3456	5.	5475
4	0.	3543	10.	325	0.	3630	0.	6899	0.	489	2.	9719	6.	3786	15.	0090	18.	6413	25.	4997
5	0.	2157	14.	7888	0.	2273	0.	7889	0.	2187	3.	3434	9.	1666	18.	1392	21.	5216	30.	8103
6	0.	967	0.	26069	0.	9664	0.	9971	0.	881	0.	9904	7.	2099	9.	9144	10.	5934	17.	3215
7	0.	6349	0.	14142	0.	6350	0.	8451	0.	634	1.	5705	1.	8527	7.	4923	10.	4385	12.	9309
8	0.	0003	28.	4174	0.	0155	0.	8960	0.	0133	5.	9982	24.	1588	35.	3920	38.	6284	59.	9426
9	0.	8596	15.	5334	0.	8561	0.	9939	0.	887	5.	3792	46.	5090	42.	9412	40.	9482	73.	5306
10	0.	2293	14.	999	0.	2194	0.	9003	0.	2095	5.	4913	18.	4368	30.	4560	34.	4040	51.	5077
11	0.	0430	0.	0761	0.	0511	0.	1489	0.	0522	0.	849	0.	1213	3.	8791	6.	3469	6.	9175
12	0.	1559	0.	0275	0.	1619	0.	3282	0.	1667	0.	0387	0.	3184	1.	8017	3.	1486	3.	5538
MMRE	4.	135%	87.	97%	4.	17%	7.	989%	4.	1697%	29.	428%	116.	55%	179.	809%	206.	39%	308.	09%

Figure 2.

MMRE and MD for proposed BBO, PSO, FPA, GA, COCOMO-II and others models using NASA datasets.

Figure 3.

Comparison among actual efforts vs BBO, PSO, FPA, GA and COCOMO-II for NASA (c) and Turkish Industry (d) datasets.

Figure 4.

MD and MMRE for proposed BBO, PSO, GA, COCOMO-II and others models using Turkish Industry datasets.

5. Analysis and discussion

Experimental work was conducted on PC with core i 7 processor 3.3 GHz and 16.00 GB Ram. The proposed algorithm is implemented in MATLAB-2013 tool. The experimental work is carried out by using NASA-93 and Turkish Industry software projects. For NASA-93 software projects, we used first 40 projects for training data and 35 randomly selected software projects for testing. For Turkish Industry software projects overall 12 projects were used. In our experiments, we obtained A $=$ 4.0237 and B $=$ 0.86120 (for proposed BBO model using NASA-93 software projects), A $=$ 4.945 and B $=$ 0.726 (for GA) while A $=$ 4.5949 and B $=$ 0.7906 using PSO algorithm. While A $=$ 4.2826, B $=-$ 0.1757 (for proposed BBO using Turkish Industry software projects), A $=$ 4.719 and B $=-$ 0.858 (for GA) while A $=$ 4.2366 and B $=-$ 0.1682 using PSO algorithm. Using both dataset NASA-93 and Turkish Industry, the proposed BBO method provides better results compared to COCOMO-II, PSO, FPA, GA, IVR, Doty, Halstead, SEl, and Bailey Basil in form of MMRE and MD as shown in Tables 7 and 8 respectively.

Table 9
MMRE (mean MMRE) and MD (Manhattan distance) comparison for proposed BBO, PSO, GA and other various models using NASA-93 datasets

Model	MMRE (mean MMRE)	MD (Manhattan distance)
COCOMO-II	39.37	12064.14
PSO	35.71	12001.41
GA	36.33	12050.02
IVR	47.84	13085.34
Doty	55.43	13159.52
Halstead	58.70	13365.92
SEL	77.67	16960.76
Baily Basil	103.55	14428.39
FPA	28.37	11840.54
Proposed-BBO	27.04	11380.59

Table 10

MMRE (mean MMRE) and MD (Manhattan distance) comparison for Proposed BBO, PSO, GA and other various models using Turkish Industry datasets

Model	MMRE (mean MMRE)	MD (Manhattan distance)
COCOMO-II	733.13	585.9424
PSO	34.800	43.3571
GA	66.57	60.0558
SEL	245.23	201.1912
Halstead	971.30	1254.72
IVR	1498.41	1459.898
Doty	1719.98	1520.708
Baily Basil	256.749	2504.08
FPA	34.54	43.3417
Proposed-BBO	34.47	43.2952

It clearly seen that the proposed BBO algorithm is able to reduce error efficiently. Moreover, the performance of PSO and GA algorithms are explained in detail for NASA-93 and Turkish Industry software projects. The Fig. 3 is the graphical representation of the effort results for both datasets. The graphs also represent that using BBO-COCOMO-II is much closer to the actual effort when compared to efforts predicted by the original parameters of COCOMO-II, optimized COCOMO-II using BBO-COCOMO-II, PSO and GA algorithms. In these figures red curve represents the proposed BBO algorithm. The Figs 2 and 4 show the MMRE and MD comparison of the proposed BBO and other cost estimation techniques using NASA-93 and Turkish Industry software projects respectively.

Moreover, Tables 9 and 10 present comparison analysis of MRE among the COCOMO-II with BBO-COCOMO-II, PSO, GA, FPA and other different cost estimation models in term of effort for 30 projects from NASA and 12 projects from Turkish Industry software projects datasets.

6. Conclusion and future recommendation

In this study, we applied a novel approach biogeography-based algorithm as a natured inspired optimization algorithm for optimizing the COCOMO-II current parameters. The proposed BBO-COCOMO-II has been tested by using NASA-93 and Turkish Industry software projects. Simulation results show that the proposed BBO-based method outperforms conventional COCOMO-II, PSO, GA, FPA, and other cost estimation techniques. In the future, we intend to use other evolutionary algorithms for optimizing the coefficients of the constructive cost model and constructive quality estimation model.

Footnotes

Acknowledgments

We thank the editor and reviewers for their thorough reviewers, thoughtful comments, and constructive suggestions.

References

Nandal

Sangwan

. Software cost estimation by optimizing cocomo model using hybrid batgsa algorithm. International Journal of Intelligent Engineering and Systems. 2018; 11(4): 250–263.

Langsari

Sarno

. Optimizing COCOMO II parameters using particle swarm method. In: 2017 3rd International Conference on Science in Information Technology (ICSITech). IEEE; 2017. pp. 29–34.

Ullah

Wang

Sheng

Long

Asim

Riaz

. A Novel Technique of Software Cost Estimation Using Flower Pollination Algorithm. In: 2019 International Conference on Intelligent Computing, Automation and Systems (ICICAS). IEEE; 2019. pp. 654–658.

Trendowicz

Jeffery

. Software project effort estimation. Foundations and Best Practice Guidelines for Success, Constructive Cost Model – COCOMO pags. 2014; 277–293.

Salam

Khan

Baseer

. A comparative study for software cost estimation using COCOMO-II and Walston-Felix models. In: The 1st International Conference on Innovations in Computer Science & Software Engineering, (ICONICS 2016). 2016. pp. 15–16.

Zhang

Yang

Wang

. A study on software effort prediction using machine learning techniques. In: International Conference on Evaluation of Novel Approaches to Software Engineering. Springer; 2011. pp. 1–15.

Grover

Bhatia

Mittal

. Estimating Software Test Effort Based on Revised UCP Model Using Fuzzy Technique. In: International Conference on Information and Communication Technology for Intelligent Systems. Springer; 2017. pp. 490–498.

Kumari

Pushkar

. Software cost estimation using cuckoo search. In: Advances in Computational Intelligence. Springer; 2017. pp. 167–175.

Kumari

Pushkar

. Cuckoo search based hybrid models for improving the accuracy of software effort estimation. Microsystem Technologies. 2018; 24(12): 4767–4774.

10.

Gharehchopogh

Rezaii

Arasteh

. A new approach by using Tabu search and genetic algorithms in Software Cost estimation. In: 2015 9th International Conference on Application of Information and Communication Technologies (AICT). IEEE; 2015. pp. 113–117.

11.

Urbanek

Prokopova

Silhavy

Kuncar

. Using analytical programming for software effort estimation. In: Computer Science Online Conference. Springer; 2016. pp. 261–272.

12.

Dalal

Dahiya

Jaglan

. Efficient Tuning of COCOMO Model Cost Drivers Through Generalized Reduced Gradient (GRG) Nonlinear Optimization with Best-Fit Analysis. In: Progress in Advanced Computing and Intelligent Engineering. Springer; 2018. pp. 347–354.

13.

Santos

Ferreira

. Safety critical software effort estimation using COCOMO II: a case study in aeronautical industry. IEEE Latin America Transactions. 2018; 16(7): 2069–2078.

14.

Ghatasheh

Faris

Aljarah

Al-Sayyed

. Optimizing software effort estimation models using firefl algorithm. arXiv preprint arXiv:190302079. 2019.

15.

Attarzadeh

Mehranzadeh

Barati

. Proposing an enhanced artificial neural network prediction model to improve the accuracy in software effort estimation. In: 2012 Fourth International Conference on Computational Intelligence, Communication Systems and Networks. IEEE; 2012. pp. 167–172.

16.

Shepperd

Schofi

. Estimating software project effort using analogies. Series on Software Engineering and Knowledge Engineering. 2005; 16: 64.

17.

Effendi

Sarno

Prasetyo

. Implementation of Bat Algorithm for COCOMO II Optimization. In: 2018 International Seminar on Application for Technology of Information and Communication. IEEE; 2018. pp. 441–446.

18.

Ahmad

Bamnote

. Whale-crow optimization (WCO)-based optimal regression model for software cost estimation. Sädhanä. 2019; 44(4): 94.

19.

Sheta

Rine

Ayesh

. Development of software effort and schedule estimation models using soft computing techniques. In: 2008 IEEE Congress on Evolutionary Computation (IEEE World Congress on Computational Intelligence). IEEE; 2008. pp. 1283–1289.

20.

Boehm

Sullivan

. Software Economics. CrossTalk, The Journal of Defense Software Engineering. 1981.

21.

Boehm

Clark

Horowitz

Westland

Madachy

Selby

. Cost models for future software life cycle processes: COCOMO 2.0. Annals of Software Engineering. 1995; 1(1): 57–94.

22.

Srivastava

Chauhan

Singh

. VRS model: a model for estimation of efforts and time duration in development of IVR software system. Int J of Software Engineering, IJSE. 2012; 5(1).

23.

Chulani

. Bayesian analysis of software cost and quality models. In: Proceedings IEEE International Conference on Software Maintenance. ICSM 2001. IEEE; 2001. pp. 565–568.

24.

Boehm

Papaccio

. Understanding and controlling software costs. IEEE Transactions on Software Engineering. 1988; 14(10): 1462–1477.

25.

Rarick

Simon

Villaseca

Vyakaranam

. Biogeography-based optimization and the solution of the power fl w problem. In: 2009 IEEE International Conference on Systems, Man and Cybernetics. IEEE; 2009. pp. 1003–1008.

26.

Yahya

Ahmad

Lee

. Effects of software process maturity on COCOMO IIs effort estimation from CMMI perspective. In: 2008 IEEE International Conference on Research, Innovation and Vision for the Future in Computing and Communication Technologies. IEEE; 2008. pp. 255–262.

27.

Manalif

. Fuzzy Expert-COCOMO risk assessment and effort contingency model in software project management. 2013.

28.

Zheng

Zhang

Chen

. Biogeography-based optimization: algorithms and applications. 2019.

29.

Alam

Khan

Asim

. Hybrid biogeography based optimization algorithm for optimization problems. Gomal University Journal of Research. 2017; 33(1): 134–142.

Optimization of software cost estimation model based on biogeography-based optimization algorithm

Abstract

Keywords

1. Introduction

4.1 Fitness function

4.2 BBO algorithm

4.3.1 Data set description

Table 3 Parameter settings in BBO algorithm

Table 9 MMRE (mean MMRE) and MD (Manhattan distance) comparison for proposed BBO, PSO, GA and other various models using NASA-93 datasets

Footnotes

Acknowledgments

References

Table 3
Parameter settings in BBO algorithm

Table 9
MMRE (mean MMRE) and MD (Manhattan distance) comparison for proposed BBO, PSO, GA and other various models using NASA-93 datasets