The evaluation metric for this competition is Categorization Accuracy. This is the number of category predictions that match the actual category, divided by the total number of observations. Take the following example:
USER_ID     Actual       Prediction
123     "Democrat"     "Democrat"
124     "Republican"   "Democrat"
130     "Republican"   "Republican"
145     "Democrat"     "Republican"
In this case, two predictions match the actual category out of four total observations. Consequently, the Categorization Accuracy in this case is 0.5.
Submission Format
For every author in the dataset
, submission files should contain two columns: USER_ID and PREDICTION. Use comma-delimited submission files.
The file should contain a header and have the following format:
USER_ID,PREDICTION
123,Democrat
124,Democrat
130,Republican
145,Republican
etc.