Skip to main content
. 2019 Oct 10;17(10):e3000416. doi: 10.1371/journal.pbio.3000416

Fig 2. Training a machine learning system to predict future translation.

Fig 2

(A) Schematic of inputs into the machine learning system. Papers were assigned the same HAMC scores used for visualization on the trilinear graphs, as well as 3 binary indicators, one each for the presence of modifying MeSH terms in either the Disease, Therapeutic/Diagnostic Approaches, or Chemical/Drug categories. These properties were also scored for papers citing the article of interest, and the citing network was summarized by the max, mean, and standard deviations (SDs), as well as the overall citation rate (cites/year). For this analysis, citation rate is preferable to RCR [18], because citations per year can be used immediately while a meaningful citation count must accrue before RCR can be calculated. (B) Schema for training the machine learning model and generating predictions. HAMC, Human, Animal, and Molecular/Cellular; max, maximum; MeSH, Medical Subject Headings; RCR, Relative Citation Ratio.