ABSTRACT: Software is a discipline that addresses all aspects of software production, starting from the early stages of capturing requirements (user needs analysis), specification (specify the specification of user needs), design, coding, testing to maintenance of the system after use [1]. Accurate estimation is the key to the success of software development projects. There are many mechanisms of the estimated costs and business software, but estimated costs and accurate business is still a challenge for researchers and managers of software development projects [2].
Research that uses neural network method in estimating the cost and effort of software development projects have error rates are still high, so it still needs improvement to obtain a more optimal result. In this study created a model of neural network algorithms and neural network algorithm model based particle swarm optimization to get the architecture in estimating the cost and effort of software development projects and to obtain a lower rate of error.
After testing the two models of neural network algorithms and neural network algorithm based on particle swarm optimization, the results obtained are neural network algorithm generates values Magnitude Mean Relative Error (MMRE) of 0965 and Prediction value of 25% or Pred (0.25) by 70% , but after the improvement of the model is based on neural network algorithm particle swarm optimization MMRE value obtained for 0668 and the value Pred (0.25) by 95%. So that both methods have different MMRE value that is equal to 0.297 and the difference in value Pred (0.25) by 25%.
This study can be enhanced by the selection of appropriate methods for several attributes such as: virt, turn, often, Aexp, PCAP, vexp, lexp and SCED that could affect attribute weights. Furthermore, comparing datasets COCOMO 81 with another dataset, to see the model and measurement results are better. This research can be developed with other optimization methods such as NN + GA, SVM + PSO, or others.
Keywords: Effort and Cost Estimation, Software Engineering, Neural Network, Particle Swarm optimization
Penulis: Ramli Ahmad, Baiq Andriska Candra Permana, Hariman Bahtiar
Kode Jurnal: jptinformatikadd170278

Artikel Terkait :