Unlocking predictive potential
Meenakshi, Meenakshi Pareek · 2024
Artificial intelligence (AI) plays a powerful role to increase the reliability, performance, and accuracy of software development. Accurate software work estimation serves as essential for successful project planning. Traditional and Agile techniques give baseline estimates but are frequently inadequate. Machine learning algorithms, such as regression and SVM, increase accuracy by detecting complicated data patterns. Ensemble models improve performance by merging many models. Deep learning (DL) excels at modelling complicated patterns but confronts optimisation and interpretability issues. To solve this, hybrid models that combine DL and optimisation approaches, such as genetic algorithms, have been created. The proposed models optimised DL architectures for software effort estimation (SEE), resulting in higher performance. Empirical results reveal hybrid DL with an optimisation model outperforms conventional, agile, solo machine learning, and independent DL approaches. Performance matrix (MAE: 9.0, RMSE: 12.5, MMRE: 0.10, PRED(25%): 95%) reveals that DL mixed with optimisation is more efficient for estimating software work than traditional, agile, solo machine learning, and solo DL approaches. This chapter covers how DL influences the software effort estimate, including the approach it uses, applications, benefits, problems, and future possibilities.