Predicting votes from wards with unknown data

Now that we know how to train our models and find the best one possible, we will provide predictions for those wards for which we don't have voting data using the best models we found using the Vote measure. To do so, we simply execute the following line:

predictions <- predict(best_lm_fit_by_votes, data_incomplete)

predictions
#>    804    805    806    807    808    809    810    811    812    813
#> 0.6845 0.6238 0.5286 0.4092 0.5236 0.6727 0.6322 0.6723 0.6891 0.6004
#>    814    815    816    817    818    819    820    821    822    823
#> 0.6426 0.5854 0.6966 0.6073 0.4869 0.5974 0.5611 0.4784 0.5534 0.6151
(Truncated output)

This will take the best model we found earlier using the Votes measure and use it to generate predictions for the Proportion variable in the data_incomplete data, which contains those observations for which we don't have any voting data. These are the best predictions we can provide with what we have...