airbnb_multicity.py

import datasets
import pyarrow as pa
import pyarrow.parquet as pq


DESCRIPTION = "The dataset contains Airbnb data from 80 capitals and major cities all around the world."
# DATA_URL="https://huggingface.co/datasets/kraina/airbnb_multicity/resolve/main/data/all_airbnb.parquet"

DATA_DIRS = ["benchmark", "all"]
RESOLUTIONS=["8","9","10"]

class AirbnbDatasetConfig(datasets.BuilderConfig):
    """BuilderConfig """

    def __init__(self, data_url, **kwargs):
        """BuilderConfig.
        Args:
          **kwargs: keyword arguments forwarded to super.
        """
        super(AirbnbDatasetConfig, self).__init__(**kwargs)
        self.data_url = data_url
        

class AirbnbDataset(datasets.ArrowBasedBuilder):
    BUILDER_CONFIG_CLASS = AirbnbDatasetConfig
    DEFAULT_CONFIG_NAME = "8"

    BUILDER_CONFIGS = [
        AirbnbDatasetConfig(
            name = res,
            description = f"This is the official train test split for Airbnb Datatset in h3 resolution = {res}. Benchmark cities are: Paris, London, Rome, Melbourne, New York City, Amsterdam.",
            data_url={
                "train": f"https://huggingface.co/datasets/kraina/airbnb_multicity/resolve/main/data/res_{res}/airbnb_train.parquet",
                "test": f"https://huggingface.co/datasets/kraina/airbnb_multicity/resolve/main/data/res_{res}/airbnb_test.parquet"
            }
        )
        for res in RESOLUTIONS
    ]

    BUILDER_CONFIGS = BUILDER_CONFIGS + [
        AirbnbDatasetConfig(
            name="all",
            description=f"This is a raw, full version of Airbnb Dataset."+DESCRIPTION,
            data_url={"train":f"https://huggingface.co/datasets/kraina/airbnb_multicity/resolve/main/data/all_airbnb.parquet"}

        )]
    
    def _info(self):
        return datasets.DatasetInfo(
            # This is the description that will appear on the datasets page.
            description=self.config.description,
            homepage="https://insideairbnb.com/",
            citation="",
            # This defines the different columns of the dataset and their types
            features=datasets.Features(
                {   "id": datasets.Value(dtype="int64"),
                    "name": datasets.Value(dtype="string"),
                    "host_id": datasets.Value(dtype="int64"),
                    "host_name": datasets.Value(dtype="string"),
                    "latitude": datasets.Value(dtype="float64"),
                    "longitude": datasets.Value(dtype="float64"),
                    "neighbourhood": datasets.Value(dtype="string"),
                    "room_type":datasets.Value(dtype="string"),
                    "price":datasets.Value(dtype="float64"),
                    "minimum_nights":datasets.Value(dtype="int64"),
                    "number_of_reviews":datasets.Value(dtype="int64"),
                    "last_review": datasets.Value(dtype="string"),
                    "reviews_per_month":datasets.Value(dtype="float64"),
                    "calculated_host_listings_count":datasets.Value(dtype="int64"),
                    "availability_365":datasets.Value(dtype="int64"),
                    "number_of_reviews_ltm":datasets.Value(dtype="int64"),
                    "city":datasets.Value(dtype="string"),
                    "date":datasets.Value(dtype="string"),
                    # These are the features of your dataset like images, labels ...
                }
            ),
        )


    def _split_generators(self, dl_manager: datasets.download.DownloadManager):
        downloaded_files = dl_manager.download(self.config.data_url)
        if self.config.name == "all":
            return [
                datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={'filepath': downloaded_files["train"]})
            ]
        else:
            return [
                    datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={'filepath': downloaded_files["train"]}),
                    datasets.SplitGenerator(name=datasets.Split.TEST, gen_kwargs={'filepath': downloaded_files["test"]})
                ]
    # method parameters are unpacked from `gen_kwargs` as given in `_split_generators`
    def _generate_tables(self, filepath):
        with open(filepath, mode="rb") as f:
            parquet_file = pq.ParquetFile(source=filepath)
            for batch_idx, record_batch in enumerate(parquet_file.iter_batches()):
                df = record_batch.to_pandas()
                df.reset_index(drop=True, inplace=True)
                pa_table = pa.Table.from_pandas(df)
                yield f"{batch_idx}", pa_table