Analysts use Athena to query years of clickstream CSV files in S3. Most queries select a few columns for one day, but currently scan almost all historical data. Which TWO changes best reduce scanned data and improve these queries? (Choose TWO.)

1. Put every file into one unpartitioned prefix and select every column in every query.
2. Convert the dataset to a suitable compressed columnar format such as Parquet.
3. Partition the data by event date and include appropriate partition filters in queries.
4. Create a Glue crawler that discovers the existing unpartitioned CSV schema but leave file layout and queries unchanged.
5. Transition all active query files to Glacier Deep Archive and query them immediately without restoration.