Treffer: Hybridizing Intelligence: A Comparative Study of Machine Learning Algorithm and ANN-PSO Deep Learning Model for Software Effort Estimation.

Software Effort Estimation (SEE) is a technique for properly anticipating staff during the development of a project. Software development process is complicated and the most critical step is estimating the staff effort required for the projects. Correctly identifying the precise quantity of effort required in the initial phases of development can be problematic. Researchers have developed many machine learning and deep learning algorithms to improve the accuracy. A hybrid model PSO (Particle swarm optimisation)-based artificial neural network (ANN) model for software effort estimating (SEE) is presented and has demonstrated to outperform traditional methods. This model was evaluated using several datasets, including China, Albrecht, Kitchenham, Desharnais, Maxwell, Kemerer and Cocomo81. The machine learning algorithms Bagging, Boosting Averaging, weighted Averaging, Stacking using RF specify algorithm, are widely used in real word applications for software development. Our experiments demonstrate that the hybrid ANN-PSO model outperforms the traditional machine learning algorithm in regards to accuracy, precision, and recall. The ANN-PSO hybrid model achieves an average accuracy compared to the ML algorithm. The results of this study highlight the potential of hybrid deep learning models in tackling complex problems, particularly those involving large datasets and high-dimensional feature spaces. The hybrid ANN-optimized PSO has shown exceptional accuracy, with consistently elevated R2-squared (R 2) values across several datasets. Furthermore, the model shows performance metrics RMSE and MAE values, implying reliable predictions. These results support the effectiveness and utility of the paradigm. The low MAE of the model indicates that it may predict software development task requirements with reasonable accuracy. Given these remarkable outcomes, the hybrid PSO-optimized ANNs model will surely be important in software development. [ABSTRACT FROM AUTHOR]