Building a crosswalk which matches Apple apps to their Google Play counterparts. There is no simple solution, because the counterpart app sometimes differ in app_name and publisher_name.
Dataset 1 consists of 500 rows with columns [platform app_id app_name publisher_name categories], Dataset 2 shows the matching pairs of Dataset 1. Dataset 3 is the test-dataset, with the same shape as Dataset 1, for which the model will make predictions.
For the feature engineering part, i'm using fuzzywuzzy and spacy to get a similarity-score of the app_names and the publisher_names. I hardcoded the mapping of the categories.
I used a random forest with no special hyperparameter as prediction model.