Energy efficient and optimized genetic algorithm for software effort estimator using double hidden layer bi-directional associative memory

In software development, it's important to have an accurate assessment of effort, cost, energy, and time in order to plan and allocate resources in the best way possible. This makes it more likely that the software will work and lowers the risk that it won't. Bi-directional associative memory is used in the suggested method to figure out how long it will take to finish a project. This effort estimator was built in MATLAB and tested, verified, and trained on a large project dataset. In this work, a genetic algorithm called Double Hidden Layers Bi-directional Associative Memory is used to make a unique model for a software estimator (DHBAM). After doing several simulations, we found that the DHBAM architecture works better than the Single Hidden Layer Feed-Forward Neural Net (SHFNN) model for getting the best results. It has also been proven with the root mean square error (RMSE) method. In a previous study, the RMSE for the network design SHFNN 16-19-1 with a learning rate of 1.01 and a momentum of 0.70 after 1,000,000 iterations was 1.49074 x 10-3. With a learning rate parameter of 1.05 and a momentum parameter of 0.6, the RMSE for the network design DHBAM 16-8-6-1 is now 1.241703 x 10-3, which drops to 1.2238 x 10-3 after 100 generations in 10,000 populations using the optimized based genetic algorithm (GA). Based on the results, it's clear that the proposed effort estimator does a better job than the ones that are already in use. Experiments show that the newly proposed optimized based genetic algorithm will help re-searchers, scientists, and businesses predict important traits more accurately and efficiently early on in the planning process.

Automated summary

In software development, accurately assessing effort, cost, energy, and time is crucial for effective resource planning. A proposed effort estimator called Double Hidden Layers Bi-directional Associative Memory (DHBAM) outperforms the Single Hidden Layer Feed-Forward Neural Net (SHFNN) model in predicting project completion time. The optimized based genetic algorithm and root mean square error (RMSE) method confirm the superiority of the DHBAM architecture.