Reads the data from the input file; allows the user to set a few additional options.
![]() |
This widget performs the same function as the File widget, except that it offers the user to set the symbols used for unknown values. More importantly, the widget can construct a new domain when reloading the file, and it can reuse attributes from other files. Consider a scenario in which we have train and test data sets in two separate files. We use two file widgets to load them, build a model from the former and test in on the latter. How does Orange know which attribute from the first data set corresponds to which attribute of the second? The question may seem absurd: the attributes in the two data sets have the same names, right? Yes, that is what Orange expects: if the two data sets have exactly the same-named attributes in the same order, these attributes are considered same and everything will work as intended. Problem arises if the two data sets' attributes differ, for instance if the test set has some additional attributes. In this case, all attributes are considered different, even those with the same name. (This may change in the future!). For such cases, you can use the Extended File widget: one of the data sets can be loaded with the ordinary file widget, and the second with the extended one. The latter should be connected to the former, which will feed him a list of attributes to reuse. |
In this typical schema, learning examples are loaded using the File widget. Descriptions of some or all of their attributes are reused by Extended File widgets which loads the test data. Examples from both files are fed to Test Learners as learning and testing examples.