SPARQL is the W3C [Prud and Seaborne, 2006] recommendation for querying datasets structured according to the Resource Description Framework (RDF) [Lassila and Swick, 1998].
We applied the radial basis function (RBF) as the kernel types in accordance with our datasets structure, where γ > 0 is the kernel parameter as shown in (4) k (x i, x j ) = e − γ | x 1 − x j | 2. Subsequently, these datasets were mapped into the linearly separable space, and hyperplanes divided them into two labeled classes.
When the numbers of datasets and variables to be harmonised are small, dataset structures can be handled manually.
The success rate was defined as the number of correctly predicted proteins divided by the total number of dataset structures.
> -wrap-foot> For the training dataset, structures aligned to the membrane normal were obtained from the OPM database (Lomize et al., 2006).
This article presents an automated approach to harmonise dataset structures, which optimises the search of variables, minimises manual inputs and reduces the risk of error.
When the number of datasets and variables are large (e.g. there were more than 250 DHS and MICS datasets, with several hundreds of variables in some of our analyses), harmonisation of dataset structures becomes tedious, as well as time-consuming and extremely prone to error.
Designs, data collection and data management procedures are relatively standard for the same type of survey (DHS and MICS); however, there are differences in dataset structures between DHS and MICS and within each of these types of surveys (e.g. diverse variables names for the same variable and/or different coding of the same variables).
As might be expected, we observe that when the intent of an HTE study moves from exploratory to confirmatory, so do the technical challenges associated with a given methodological approach – in terms of more complex dataset structures, assumptions related to distributions, or sample size considerations.
Fig. 8 The dataset structure of the SciData Data Model.
