A genetic algorithm based framework for software effort prediction

Abstract

Background: Several prediction models have been proposed in the literature using different techniques obtaining different results in different contexts. The need for accurate effort predictions for projects is one of the most critical and complex issues in the software industry. The automated selection and the combination of techniques in alternative ways could improve the overall accuracy of the prediction models. Objectives: In this study, we validate an automated genetic framework, and then conduct a sensitivity analysis across different genetic configurations. Following is the comparison of the framework with a baseline random guessing and an exhaustive framework. Lastly, we investigate the performance results of the best learning schemes. Methods: In total, six hundred learning schemes that include the combination of eight data preprocessors, five attribute selectors and fifteen modeling techniques represent our search space. The genetic framework, through the elitism technique, selects the best learning schemes automatically. The best learning scheme in this context means the combination of data preprocessing + attribute selection + learning algorithm with the highest coefficient correlation possible. The selected learning schemes are applied to eight datasets extracted from the ISBSG R12 Dataset. Results: The genetic framework performs as good as an exhaustive framework. The analysis of the standardized accuracy (SA) measure revealed that all best learning schemes selected by the genetic framework outperforms the baseline random guessing by 45-80%. The sensitivity analysis confirms the stability between different genetic configurations. Conclusions: The genetic framework is stable, performs better than a random guessing approach, and is as good as an exhaustive framework. Our results confirm previous ones in the field, simple regression techniques with transformations could perform as well as nonlinear techniques, and ensembles of learning machines techniques such as SMO, M5P or M5R could optimize effort predictions.

Estimation Approaches of Machine Learning in Scrum Projects: A Review

“…Distinctive estimation settings are made with reference to in writing, for example, little informational collection, anomalies, absolute highlights, and missing qualities. Analysts recommend [21,22] that it is more productive to decide the best model in a specific setting instead of deciding the best single model, since estimate models carry on uniquely in contrast to one dataset to other, which makes them precarious. Studies directed on information mining report that group strategies furnish exact outcomes in examination with single strategies as every strategy has quality and shortcoming so joining will moderate the shortcoming.…”

“…Studies directed on information mining report that group strategies furnish exact outcomes in examination with single strategies as every strategy has quality and shortcoming so joining will moderate the shortcoming. Outfit effort estimation systems might be gathered into two noteworthy classes [22][23][24]16]: homogeneous (e.g., bagging and SVR, RF, MLP, LR, RBF, ANFIS, CBR, RF, SGB, CART, and so forth) and heterogeneous perceived by their base models and blend rules. ANN was utilized most with outfits.…”

Threats to validity in search‐based predictive modelling for software engineering

“…These performance metrics are specific to the software attribute being addressed. It could be classification accuracy, precision, recall and so on for defect prediction models or mean magnitude [17] 14.5 (28.0) S35: Minku [48] 12 (3.3) S66: Jain [79] 10.5 (1.0) S5: Malhotra [18] 14.5 (*) S36: Malhotra [49] 12 (13.3) S67: Dolado [80] 10 (13.2) S6: Hosseini [19] 14.5 (*) S37: Malhotra [50] 12 (1.3) S68: Singh [81] 10 (0.6) S7: De Carvalho [20] 14 (9.0) S38: Shukla [51] 11.5 (6.6) S69: Li [82] 10 (10.4) S8: Canfora [21] 14 (16.0) S39: Liu [52] 11.5 (3.2) S70: Sheta [83] 10 (2.0) S9: Abdi [22] 14 (0.5) S40: Kirsopp [53] 11.5 (8.7) S71: Araujo [84] 10 (3.6) S10: Murrillo-Morera [23] 14 (*) S41: Shan [54] 11.5 (4.4) S72: Malhotra [85] 10 (0.5) S11: Ferrucci [24] 13.5 (2.0) S42: Ferrucci [55] 11.5 (0.7) S73: Malhotra [86] 10 (1.7) S12: Oliveira [25] 13.5 (14.9) S43: Bardsiri [56] 11.5 (11.5) S74: Kumar [87] 10 (1.0) S13: Bardsiri [26] 13.5 (5.7) S44: Basgalupp [57] 11.5 (6.3) S75: Li [88] 9.5 (20.4) S14: Arar [27] 13. of relative error (MMRE) while developing effort estimation models. This specific characteristic makes SBAs ideal for developing SPMs as they can find a good solution by optimising these performance metrics employed as fitness functions (S1, S5, S15, S25, S34, S45).…”

“…All the studies whose QS was <7.5 (50% of the total QS) were rejected. After this step, a total of 93 literature studies [14–106] were selected, which were termed as primary studies of our review. Relevant data pertaining to RQs was extracted from these studies and the obtained results are reported in Section 4.…”

Statistical Testing on Prediction of Software Defects

“…Juan Murillo Morera designed a frame work for software effort prediction using genetic algorithm. The performance of learning schemes was measured using the metrics Spearman's rank correlation, mean of magnitude relative error, median of magnitude of relative error, standardized accuracy, number of predictions within percentage of actual ones [7]. J S Pahariya et al proposed Genetic Programming based feature selection for software cost estimation.…”