Code Complexity, Size, and Contributors as Determinants of Software Bugs

Abstract

Software reliability plays a crucial role in ensuring that development activities run effectively, and one of its key indicators is the number of bugs identified during the development process. This study explores how code complexity, code size, and the number of contributors serve as determinants of bug occurrence by applying a multiple linear regression approach. The analysis used 32 software projects sourced from the 2024 Software Engineering & Code Quality Dataset, all of which satisfied classical assumption tests. The resulting model demonstrates strong explanatory power with an R² value of 0.863. The F-test confirms that the three determinants jointly influence bug frequency in a statistically significant manner (F = 58.847 > F critical = 2.9467; p = 0.000). Based on partial t-tests, code size emerges as the only significant determinant (t = 5.675; p = 0.000), whereas code complexity and contributor count do not show meaningful partial effects. The regression model obtained is Ŷ = −30.968 + 5.308X₁ + 0.000X₂ + 1.068X₃. These findings suggest that efforts to reduce software bugs should primarily focus on managing the growth of code size.