Extracted model tool
|Moderator:Nobody (see all)|
Click here to sign up.
Extracted model is a computing model that is created by extracting code out of a set of variables within the integrated resource platform. The code is an operationalisation of the definition of the variable. The code is located in the Definition/Formula attribute within an appropriate tag. For example, Analytica code is between <anacode>place code here</anacode> tags; R code is between <rcode>place code here</rcode> tags. When the code is compiled, it creates a fully functional model that can be run.
Model extractor is a tool that extracts the computer code out of the variable definition/formula attributes and compiles a self-sustained model file from them. The model can then be run by an external software, and the results can be uploaded to the result database.
- Works only for simple, examplary assessments
- For all variables models are needed
- The causality is represented in a variable via the attribute formula: a complex model with 1000s of lines of code working in timesteps and on different scales cannot be integrated into a variable -> e.g. dispersion models, usually Eulerian models, cannot be integrated into a model extractor; for these, external models are used by applying the variable transfer protocol.
Structure of the process
Extracted models are flexible in the sense that there is no need to restrict ourselves to a single software. Any object-oriented programming code can be attached to a group of variables, and those variables can be calculated using the particular software independently of the other variables and their calculation software. One variable can even contain several codes for different software, although this is not recommended due to possible updating problems. However, this is only possible when:
- The calculation software used utilises object-oriented programming,
- The software is able to take the data for the input variables (the most upstream variables that cannot be calculated inside the extracted model) from the result database.
However, parts of this information flow can, at least in the preliminary phase, be done manually.
Possible software that may have the required functionalities:
- Analytica, a Monte Carlo simulation program with graphical interface.
- R, an open access statistical software
- Unicorn, the "Uncertainty analysis with Correlations" (UNICORN) software tool, implementing staff research work on dependence modelling for high dimensional distributions.