EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Assessment of accuracy for deep learning and other machine learning as binary classifier algorithms

Theophilus Ben-Bala, Abubakar Yahaya, Aminu Suleiman Mohammed, Reuben Oluwabukunmi David, Aliyu Ismail Ishaq

Journal of Statistical Sciences and Computational Intelligence · 12.06.2026 · 2 (2) · 559-575

DOI: 10.64497/jssci.90 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Statistical Sciences and Computational Intelligence, online erschienen am 12.06.2026. Verfasst von Theophilus Ben-Bala und 4 weiteren Autor:innen. Die Arbeit nennt 34 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Machine Learning (ML) and Deep Learning (DL) are widely employed in binary classification tasks across various domains. While DL models are often assumed to outperform ML models, this study conducts a statistical assessment to determine whether this superiority is statistically significant. Using twelve datasets from diverse domains, three ML algorithms- Regression, K-Nearest Neighbors (KNN), and Naïve Bayes- were compared against a Multi-Layer Perceptron (MLP) model representing DL. The study also incorporates accuracy results from nineteen prior research studies. A rigorous preprocessing pipeline was employed, including feature selection using recursive feature elimination (RFE), Chi-square tests, and variance threshold methods. Models were implemented in Python, trained on 80% of each dataset, and tested on the remaining 20%. To evaluate performance differences, an independent samples t-test was conducted after validating assumptions of normality (Shapiro-Wilk test) and equal variance (Levene’s test). The results show that while DL models exhibited slightly higher average accuracy, the observed difference was not statistically significant (p = 0.517 for fitted models and p = 0.131 for literature-based results). This suggests that ML models can achieve comparable performance to DL models in binary classification tasks, challenging the notion that DL is inherently superior. The findings highlight the importance of algorithm selection based on computational efficiency, dataset characteristics, and application requirements, rather than an assumption of DL dominance. This study contributes to the ongoing debate on ML versus DL for binary classification, providing empirical evidence that both approaches can be effective depending on context. Future research should explore larger datasets, alternative performance metrics, and domain-specific optimizations to further substantiate these findings.

Zitieren

APA

Ben-Bala, T., Yahaya, A., Mohammed, A. S., David, R. O., & Ishaq, A. I. (2026). Assessment of accuracy for deep learning and other machine learning as binary classifier algorithms. Journal of Statistical Sciences and Computational Intelligence, 2(2), 559-575. https://doi.org/10.64497/jssci.90

Vancouver

Ben-Bala T, Yahaya A, Mohammed AS, David RO, Ishaq AI. Assessment of accuracy for deep learning and other machine learning as binary classifier algorithms. Journal of Statistical Sciences and Computational Intelligence. 2026;2(2):559-575. doi:10.64497/jssci.90

BibTeX
@article{BenBala2026,
  author = {Ben-Bala, Theophilus and Yahaya, Abubakar and Mohammed, Aminu Suleiman and David, Reuben Oluwabukunmi and Ishaq, Aliyu Ismail},
  title = {Assessment of accuracy for deep learning and other machine learning as binary classifier algorithms},
  journal = {Journal of Statistical Sciences and Computational Intelligence},
  year = {2026},
  volume = {2},
  number = {2},
  pages = {559--575},
  doi = {10.64497/jssci.90}
}
Kostenlos starten →