Browsing by Author "Saeidmehr, Amirhossein"

Now showing 1 - 2 of 2
  • Thumbnail Image
    ItemOpen Access
    Systematic Review using a Spiral approach with Machine Learning
    (2023-03-22) Saeidmehr, Amirhossein; Samavati, Faramarz; Steel, Piers; Maleki, Farhad; Chapman, Derek
    Systematic reviews have become increasingly time-consuming and costly due to the accelerating growth of academic literature, doubling every nine years. Machine learning (ML) offers a promising solution to manage the growing corpus of literature, but current approaches still rely on a sequential, two-staged process designed for a purely human approach. In this thesis, we propose and test a spiral or oscillating approach, where full-text screening is done intermittently with title/abstract screening. We examine this approach in three datasets by simulating 360 conditions with different algorithmic classifiers, feature extractions, prioritization rules, data types, and information provided. Our results overwhelmingly support the spiral processing approach with Logistic Regression, TF-IDF for vectorization, and Maximum Probability for prioritization, demonstrating up to a 90\% improvement over previous two-staged ML methodologies over just title-screening, particularly for databases with fewer eligible articles. These advancements have the potential to make systematic review screening functionally achievable for another one to two decades.
  • Thumbnail Image
    ItemOpen Access
    Systematic review using a spiral approach with machine learning
    (2024-01-17) Saeidmehr, Amirhossein; Steel, Piers David Gareth; Samavati, Faramarz F.
    Abstract With the accelerating growth of the academic corpus, doubling every 9 years, machine learning is a promising avenue to make systematic review manageable. Though several notable advancements have already been made, the incorporation of machine learning is less than optimal, still relying on a sequential, staged process designed to accommodate a purely human approach, exemplified by PRISMA. Here, we test a spiral, alternating or oscillating approach, where full-text screening is done intermittently with title/abstract screening, which we examine in three datasets by simulation under 360 conditions comprised of different algorithmic classifiers, feature extractions, prioritization rules, data types, and information provided (e.g., title/abstract, full-text included). Overwhelmingly, the results favored a spiral processing approach with logistic regression, TF-IDF for vectorization, and maximum probability for prioritization. Results demonstrate up to a 90% improvement over traditional machine learning methodologies, especially for databases with fewer eligible articles. With these advancements, the screening component of most systematic reviews should remain functionally achievable for another one to two decades.