WOAMM IMPLEMENTATION ON FLANN TRAINING TO IMPROVE PERFORMANCE ON SOFTWARE EFFORT ESTIMATION

• Main Author: Aunurrahim, Fakhri
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/75283
• Institution: Institut Teknologi Bandung

Context: Software Effort Estimation (SEE) is an umbrella term covering cost, re-lated to the cost of software components. SEE is carried out to get an estimation of the cost of those components during the planning phase. The cost estimation is related to the budgeting of the software project. The expected estimation is a decent estimation that doesn’t lead to overestimation or underestimation to satisfy the actual cost within the budgeting. Related SEE researches are carried out to yield methods that are increasing the accuracy of the estimated cost of software projects. One of the method is using Functional Link Artificial Neural Network (FLANN). In accordance to machine learning—precisely Artificial Neural Network (ANN)—indicated a gap to continuously improve the learning ability. Objective: The research objective is to produce a FLANN model with Metaheuristic Algorithm (MA) that is expected to yield the learning error rate smaller than other existing models. Method: Implementing other MA, which is Whale Optimization Algorithm Modified Mutualism Phase (WOAmM). The proposed model is then compared with other existing FLANN models with MA optimization, and a model without MA op-timization as the baseline. All models are then evaluated with MAR and SA. Lastly, the evaluation results are tested with Friedman Test and Nemenyi Test posthoc. Re-sult: The proposed model yields greater accuracy than the baseline. The model also yields greater accuracy than other existing models. Conclusion: The proposed model yield better performance than other models. The model is also proven to be superior than other models in terms of statistical significance.