Noise has been acknowledged as being difficult to determine in respect of ESE datasets (Liebchen et al. 2007), especially when those datasets are secondary sources, meaning the re- searchers may be far removed from their origin. Since it is difficult to be certain about noise in a dataset, and given that researchers may be willing/able to tolerate a certain degree of noise, the assessments undertaken in this study should be interpreted as a guide to the potential state of the datasets rather than definitive statements that a dataset is noisy or otherwise. Even indicative noise assessments such as these are necessary, however, so that researchers and estimators are at least aware of the nature of the datasets they are using and can consider whether preprocessing might be beneficial in improving the quality of the data (and hence any models developed using that data). Following prior research, we employed two different approaches in determining noise for the 13 datasets selected here. The first approach was to examine whether any formulas used in deriving data were incorrect or violated relational integrity constraints (Shepperd et al. 2013), which are the stated rules/formulas or the expected outcome of a computation. The second technique utilized data classification, where incorrect classification represents a proxy for noisy instances in the data, as implemented by Liebchen et al. (2007). Classification algorithms are able to segment data into the required categories—in this study it is expected that data will be classified as “noisy” or “not noisy.” Specifically, for softwareeffortestimation, the classification algorithm identifies a record as noisy where the predicted dependent value of the classifier is different from the actual value. We used a decision tree algorithm (specifically the C4.5 algorithm available as part of the Weka data-mining toolbox) first because it is able to build relationships between data as well as to build models independent
Abstract— Accurate softwareeffortestimation is crucial for the software development process. Neither over estimated nor underestimated effort is welcome. Academic researchers and practitioners have long been searching for more accurate estimation models or methods. The objective of this paper is to explore the studies on softwareeffortestimation accuracy available in the literature and analyze the findings to obtain the answer for the question: which softwareeffortestimation model is more accurate? There were very limited reports that satisfied the research criteria. Only 8 studies with 10 reports were discovered for the analysis. It is found that Use Case Point Analysis outperforms other models with the least weighted average Mean Magnitude of Relative Error (MMRE) of 39.11%, compare to 90.38% for Function Point Analysis and 284.61% for COCOMO model. It indicates that there is still need to improve the estimation performance but the question is how.
In our work we compare the results of two different effortestimation methods and found an optimal neural network topology with best estimation method for softwareeffortestimation. In this paper we used the risk analysis to increase the accuracy of the effortestimation. After the risk will be evaluated, we have to analyze that the overall percentage of the risk will be whether acceptable, reduce or mitigate. Finally, we conclude that the present work provides accurate forecasts for the software development effort. Further work can be done by the mitigation of risk by using mitigation strategies and contingency plan when estimating the effort. And also in future we can implement the other artificial intelligence techniques and compare the performance of those techniques.
Software development effortestimation is considered a fundamental task for software develop- ment life cycle as well as for managing project cost, time and quality. Therefore, accurate estima- tion is a substantial factor in projects success and reducing the risks. In recent years, software ef- fort estimation has received a considerable amount of attention from researchers and became a challenge for software industry. In the last two decades, many researchers and practitioners pro- posed statistical and machine learning-based models for softwareeffortestimation. In this work, Firefly Algorithm is proposed as a metaheuristic optimization method for optimizing the parame- ters of three COCOMO-based models. These models include the basic COCOMO model and other two models proposed in the literature as extensions of the basic COCOMO model. The developed estimation models are evaluated using different evaluation metrics. Experimental results show high accuracy and significant error minimization of Firefly Algorithm over other metaheuristic optimization algorithms including Genetic Algorithms and Particle Swarm Optimization.
[4] compared software development effort using Fuzzy Triangular Membership Function and GBell Membership Function with COCOMO. The results were analyzed using different criterions like VAF, MARE, VARE, MMRE, Prediction and Mean BRE. Ashita Malik et al[5]. ,explored a soft computing technique to overcome the uncertainty and imprecision in software cost estimation. They used fuzzy logic in improving the effortestimation accuracy using COCOMO II by characterizing inputs parameters using Gaussian, trapezoidal and triangular membership functions and comparing their results. Ravishankar. S et al., utilize an adaptive fuzzy logic model to improve the accuracy of software time and cost estimation by using different member functions available in fuzzy logic. Ziauddin et al.[6] utilized a fuzzy logic model to improve the accuracy of softwareeffortestimation and results are compared with COCOMO II and Alaa Sheta Model.
Softwareeffortestimation is, as its name suggests, the task of estimating the amount of effort required to develop new software. Knowing the estimated effort of particular software project early in the development cycle is a valuable asset. Industry and academia have always considered a reliable and accurate estimate a challenging task. However, a review of estimation surveys by [1] documents that still less progress has been made in the area of estimation performance. Thus, there is a high demand for more research on the topic of effortestimation in software development projects. Estimators use different methods for estimation. They employ a single technique (formal model or expert judgement) or both the techniques for estimation. Passing and Shephard [2] advocated that expert judgement is leading estimation method adopted by organisations. But still it is unpredictable to define whether expert judgement is better or
softwareeffortestimation. Artificial neural network can model complex non premier relationship and approx any major able function. The most commonly adopted architecture for estimating softwareeffort is feed forward multilayer perceptron with back propagation learning algorithm and sigmoidal activation function. In this paper, it is concerned with constructive cost estimation model based on ANN. Particularly multi-layer feed forward neural network. The network model is designed accordingly to accommodate the COCOCMO model and its parameter and back propagation learning algorithm is used to train the network by iteratively processing a set of training samples and comparing the network prediction with the actual. The difference in estimation is propagated to the input for adjusting the coefficient. The process consist of repeatedly feeding input and output data from empirical observation ,propagating the error value and adjusting the connection weights until the error values fall below a user specified tolerance level.
The MMRE (Mean Magnitude Relative Error) is calculated to indicate the relative amount by which the esti- mated effort is an under-estimate or over-estimate in comparison to the actual estimate. MMRE is used in most of the research work as evaluation criterion due to its independent-of-units characteristic which means that MMRE is independent of units of estimated effort like person-hours, person-months or man-hours etc. MMRE is a meaningful tool used to summarize statistics and is very important in evaluating a softwareeffortestimation model. The MMRE values for UCP, e-UCP and Re-UCP are graphically shown in Figure 8.
The results of this benchmarking study partially confirm the results of previous studies . Simple, understandable tech- niques like OLS with log transformation of attributes and target, perform as good as (or better than) non-linear tech- niques. Additionally, a formal model such as Cocomo per- formed at least equally good as OLS with log transformation on the Coc81 and Cocnasa data sets. These two data sets were collected with the Cocomo model in mind. However, this model requires a specific set of attributes and cannot be applied on data sets that do not comply with this re- quirement. Although the performance differences can be small in absolute terms, a minor difference in estimation performance can cause more frequent and larger project cost overruns during software development. Hence, even small differences can be important from a cost and opera- tional perspective .Another conclusion is that the selection of a proper estimation technique can have a significant im- pact on the performance. A simple technique like regression is found to be well suited for softwareeffortestimation which is particularly interesting as it is a well documented technique with a number of interesting qualities like statis- tical significance testing of parameters and stepwise analy- sis. This conclusion is valid with respect to the different me- trics that are used to evaluate the techniques. Furthermore, it is shown that typically a significant performance increase can be expected by constructing softwareeffortestimation models with a limited set of highly predictive attributes. Hence, it is advised to focus on data quality rather than col- lecting as much predictive attributes as possible. Attributes related to the size of a software project, to the development, and environment characteristics, are considered to be the most important types of attributes.
[9] P. Braga, A. Oliveira, and S. Meira, “A GA-based Feature Selection and Parameters Optimization for Support Vector Regression Applied to SoftwareEffortEstimation,” in Proceedings of the 2008 ACM Symposium on Applied Computing, 2008, pp. 1788–1792. [10] G. Hu, L. Hu, H. Li, K. Li, and W. Liu, “Grid Resources Prediction With Support Vector
The main aim of any software development organizations is to finish the project within acceptable or customary schedule and budget. Budget is mainly driven by labor cost and time and together they form a measure called effort. From quality point of view estimating effort is one of the major important factors. Because estimation either it be over estimate or under estimate, produces worst results. In case of over estimation of time and effort project completion is too late due to lack of resources, which refuses the management to approve that favored project. On the other hand, under estimation may result in approval of projects that will fail to deliver the expected product within the time and budget available [11]. So, there is a need of accurate estimationeffort technique at early Stages of software development. In this research, the main aim is to improve softwareeffortestimation by using Bayesian network with PSO. The main reason for using such a learning system for this problem is to keep the estimation process up-to-date by incorporating up-to-date project data. At last Comparison is drawn between training algorithms used in
Abstract— Deep learning is an arm of Artificial Intelligence that uses deep neural networks to achieve artificial intelligence. It has made its mark in computer vision, speech recognition, language processing, and automatic engines. Google made a significant contribution to AI technologies by releasing TensorFlow (TF), its proprietary AI platform, in 2015, as an open-source software library to define, train and deploy learning models, including Machine Learning and Deep Learning. In this study, we aim to improve softwareestimation using the most recent deep learning paradigms. We employ TensorFlow and a high-level wrapper API to TF and evaluate a composite hyper-parameter tuning method employing the Cartesian grid and random search. We observe significant performance improvement, achieved (29.8%) from the base model, using the hybrid hyper-parameter tuning methodology. However, even While literature reports significant performance in cognitive imaging with TF and Keras, we have not been able to validate any substantial improvement in prediction, in the case of a softwareeffortestimation data such as ISBSG 2018 by employing these techniques. Index Terms— SoftwareEffortEstimation, Software Cost Estimation, Effort Prediction, Feature Engineering, Artificial intelligence, Neural Networks, Deep Learning, Machine Learning, TensorFlow, Keras, Hyperparameters.
Abstract: Exactness in software cost and effortestimation is tedious task in software development. Too many variables-human, technical, environmental and political-can affect the ultimate cost of software and effort applied to develop it. However, software project estimation can be transformed from a black art to a series of systematic steps that provide estimates with acceptable risk. Four module assignment factors, team size, concurrency, intensity and fragmentation were identified as potentially significant determinants of software development effort. This system is the generalized system, which can be used in any type of software development industry. The system has been developed for software development industry to reduce the scheduling effort for developing software. The results of the study indicate the work assignment factors (team size and concurrency) can be used to improve the predictive ability of effortestimation.
The above notion is supported from the survey results - 76% of the respondents agreed that the project classification by COCOMO also matters for projects developed using Object Oriented languages [10]; and because UCP model is for object oriented languages therefore the project classification of COCOMO would be considered for UCP model as well. COCOMO classifies projects into three classes to calculate the Effort with the following formula:
Software engineers try to estimate effort & cost with an accuracy in the software industry. The major target of the software engineering community is to develop useful models that can explain precisely predict the effort. There are many empirical equation based softwareestimation models had been developed over the last two- four decades , that were based on efforts estimation like Jones and Software Productivity Research‟s[2], Checkpoint model, Putnam and Quantitative Software Measurement‟s [8], SLIM model, Park[1] and PRICE Systems‟ PRICE-S model, Jensen and the SEER SEM model, Rubin and the Estimacs model and Boehm and the COCOMO model [Putnam, 1992, Jones, 1997, Park, 1988, Jensen, 1983, Rubin, 1983, Boehm, 1981, Boehm, 1995, Walkerden, 1997, Conte, 1986, Fenton, 1991, Masters, 1985, Mohanty, 1981]. These approaches impose a few restrictions, often violated by software engineering data and resulted in the development of inaccurate empirical models that do not perform very well when used for prediction of effort. This paper focuses on approximate effortestimation with the help of equation which is related to kilo line of codes (KLOC) and fuzzy multiplier.
Prof. P.K.Suri, Dean (R & D), Chairman & Professor (CSE/IT/MCA) of HCTM Technical Campus, Kaithal, since Nov. 01, 2012 . He obtained his Ph.D degree from Faculty of Engineering, Kurukshetra University, Kurukshetra and Master’s degree from IIT Roorkee (formerly . He started his research carrier from CSIR Research Scholars, AIIMS. He worked former as a dean Faculty of Engineering & Technology, Kurukshetra University, Kurukshetra, Dean Faculty of Science, KUK, Professor & Chairman of Department of Computer Sc. & Applications, KUK. He has approx 40 yrs experience in different universities like KUK, Bharakhtala University Bhopal & Haryana Agricultural university , Hissar. He has supervised 18 Ph.D. students in Computer Science and 06 students are working under his supervision. Their students are working as session judge, director & chairpersons of different institute. He has around 150 publications in International/National Journals and Conferences. He is recipient of 'THE GEORGE OOMAN MEMORIAL PRIZE' for the year 1991-92 and a RESEARCH AWARD –“The Certificate of Merit – 2000”for the paper entitled ESMD – An Expert System for Medical Diagnosis from INSTITUTION OF ENGINEERS, INDIA. The Board of Directors, governing Board Of editors & publication board of American Biographical institute recognized him with an honorary appointment to Research board of Advisors in 1999. M.P. Council of Science and Technology identified him as one of the expert to act as Judge to evaluate the quality of papers in the fields of Mathematics/Computer Science/ Statistics and their presentations for M. P. Young Scientist award in Nov. 1999 and March 2003. His teaching and research activities include Simulation and Modeling, Software Risk Management, Software Reliability, Software testing & Software Engineering processes, Temporal Databases, Ad hoc Networks, Grid Computing, and Biomechanics.
ABSTRACT: Effortestimation is one of the biggest problems faced by software industry. In software planning estimation of the effort is one of the most critical responsibilities. It is necessary to have good effortestimation in order to conduct well budget. The accuracy of the effortestimation of software projects is vital for the competitiveness of software companies. For the forecasting of softwareeffort, it is important to select the correct softwareeffortestimation techniques. Inaccurate effortestimation can be risky to an IT industry’s economics and certainty due to poor quality or trait and stakeholder’s disapproval with the software product. This paper presents M5P decision tree Technique, for effort evaluation in the field of software development.
Abstract: Accurate effortestimation of software is the state of the art problem. Setting apart, organization of software project effort, cost and time are fundamental of any software advance way. Checking for the software development in terms of size is an issue, since models such as Cost Constructive Model (COCOMO) are already present but still failure of software known as crisis is vigorous. Softwareeffortestimation if can be estimated in advanced then cost can be predicted and success or failure of software project can be determined in advance. This paper proposed a mechanism by altering the environmental variables within COCOMO model for easy and early identification of cost and effort associated with software project. Time of development is a key issue which, each and every model failed to identify. This paper proposes an additional variable consideration including electricity failure(E_F), Machine failure(M_F) and weather condition(W_C). Considering these parameters, size of organic, semidetached and embedded projects are determined. Schedule although is increased by said mechanism but accuracy is also enhanced.
Abstract – The most important activity in software project management process is the estimation of Software development effort. The literature shows many algorithmic cost estimation models such as Boehm’s COCOMO, Albrecht's Function Point Analysis, Putnam’s SLIM, ESTIMACS, Soft computing based techniques etc., but each model have their own advantages and disadvantages in predicting development cost and effort. This is because of the availability of project data in the initial stages of development process is often incomplete, inconsistent and vague. The accurate effortestimation in software project management process is major challenge. This paper is a systematic reviewof classic and contemporary literature on softwareeffortestimation. A systematicsearch is done across data sources to understand the issues and research problems ineffort estimation problem domain.