Citation

Eidi, Kamran and Subhan, Fazli and Zafarullah, Muhammad and Haider, Sajjad and Su’ud, Mazliham Mohd and Sajid, Amna (2026) Beyond prediction: Comparing cost-optimization and explainable AI for perceived success factors in software technology startups. Array, 30. p. 100945. ISSN 25900056

Text 3.pdf - Published Version Restricted to Repository staff only Download (5MB)

Abstract

Context: The Software Technology Startups (STS) run in a very uncertain and resource-constrained environment, where identification of the factors influencing the software project is a key factor to build a sustainable and successful software project. Objective: This study investigates comparative framework that combines both the cost optimization approach using the Genetic Algorithm (GA) and the predictive interpretation approach by Explainable Artificial Intelligence (XAI) for the analysis of success-oriented factors in the STS environment from a perceived perspective. Methods: A survey dataset of 147 practitioners was collected, to evaluate 14 important STS factors across the four areas of scalability, algorithms, user experience, resources, architecture, testing and technical expertise. Two probabilistic models were integrated with GA: Naïve Bayes Classifier (NBC) and K-Nearest Neighbor (KNN), with the aim of determining the cost-effective factor configuration that maximizes the perceived implementation efficacy while minimizing the normalized implementation cost. Five machine learning models (Random Forest, Support Vector Machine, XGBoost, LightGBM, and Gradient Boosting) were then used in conjunction with SHAP, permutation importance and Random Forest feature importance analysis to identify statistically significant influences on perceived success-orientation profiles. Result: The results show that there are some partial overlap but meaningful differences between optimization and prediction-based importance rankings. Under cost constraints, factors selected for GA-based optimization were User Experience (UX), Resources, and Specialized Skills while, in the XAI-based analysis, the most powerful predictive factors were Standardized Tools, Architectural Patterns and Resources. In view of these results, it is observed that the importance of the factors depends on the objective in STS environments and depends on the optimization constraints and interpretability perspectives. Conclusion: This predictive analysis is based on an engineered surrogate label based upon practitioner assessment, not on business outcomes. So, results should be seen as exploratory and perception-based, not as causal measures of startup success. The suggested framework offers an overview of decisions for practitioners looking at a balance between implementation cost, technical priorities, and perceived strategic significance, in a software start-up environment.

Item Type:	Article
Uncontrolled Keywords:	User experience (UX), Genetic algorithm, Explainable AI (XAI
Subjects:	Q Science > QA Mathematics > QA71-90 Instruments and machines > QA75.5-76.95 Electronic computers. Computer science > QA76.75-76.765 Computer software
Divisions:	Faculty of Computing and Informatics (FCI)
Depositing User:	Ms Rosnani Abd Wahab
Date Deposited:	30 Jun 2026 03:06
Last Modified:	30 Jun 2026 03:06
URII:	http://shdl.mmu.edu.my/id/eprint/16121

Downloads

Edit (login required)