geolife.py

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

_URLS = {"geolife": "https://huggingface.co/datasets/kraina/geolife/resolve/main/data/geolife.parquet"}

DESCRIPTION = "This GPS trajectory dataset was collected in (Microsoft Research Asia) Geolife project by 182 users in a period of \
    over five years  (from April 2007 to August 2012). A GPS trajectory of this dataset is represented by sequence of time-stamped points  \
        each of which contains the information of altitude, longitude, latitude. This dataset contains 17,784 trajectories, ~25M Points \
            with a total distance of 1,292,951 kilometers and a total duration of 50,176 hours. \
            These trajectories were recorded by different GPS loggers and GPS phones, and have a variety of sampling rates. \
                91.5 percent of the trajectories are logged in a dense representation, e.g. every 1~5 seconds or every 5~10 meters per point.\
                    latitude: Latitude in decimal degrees.\
                    longitude: Longitude in decimal degrees.\
                    altitude: Altitude in feet (-777 if not valid).\
                    time: Date and time as a string.\
                    mode: Way of transportation e.g. walk, taxi.\
                    trajectory_id: ID of trajectory that Point belongs to.\
                    user_id: ID of user that reported Point. \
                    crs: WGS 84"

CITATION = "[1] Yu Zheng, Lizhu Zhang, Xing Xie, Wei-Ying Ma. Mining interesting locations and travel sequences from GPS trajectories. In \
    Proceedings of International conference on World Wild Web (WWW 2009), Madrid Spain. ACM Press: 791-800.\n\
        [2] Yu Zheng, Quannan Li, Yukun Chen, Xing Xie, Wei-Ying Ma. Understanding Mobility Based on GPS Data. In Proceedings of \
            ACM conference on Ubiquitous Computing (UbiComp 2008), Seoul, Korea. ACM Press: 312-321.\n\
                [3] Yu Zheng, Xing Xie, Wei-Ying Ma, GeoLife: A Collaborative Social Networking Service among User, \
                    location and trajectory. Invited paper, in IEEE Data Engineering Bulletin. 33, 2, 2010, pp. 32-40."


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


class GeolifeDataset(datasets.ArrowBasedBuilder):
	BUILDER_CONFIG_CLASS = GeolifeDatasetConfig
	DEFAULT_CONFIG_NAME = "HMP"
	BUILDER_CONFIGS = [
		GeolifeDatasetConfig(
			name="all",
			description="GPS trajectories that were collected in (Microsoft Research Asia) Geolife project by 182 users",
			data_url={"train":"https://huggingface.co/datasets/kraina/geolife/resolve/main/data/raw/raw.parquet"}
		),
		GeolifeDatasetConfig(
			name="TTE",
			description="Official train-test split of Geolife dataset for Travel Time Estimation task. GPS trajectories that were collected in (Microsoft Research Asia) Geolife project by 182 users",
			data_url={"train":"https://huggingface.co/datasets/kraina/geolife/resolve/main/data/tte/train.parquet",
			 "test":"https://huggingface.co/datasets/kraina/geolife/resolve/main/data/tte/test.parquet"}
		),
		GeolifeDatasetConfig(
			name="HMP",
			description="Official train-test split of Geolife dataset for Human Mobility Prediction task. GPS trajectories that were collected in (Microsoft Research Asia) Geolife project by 182 users",
			data_url={"train":"https://huggingface.co/datasets/kraina/geolife/resolve/main/data/hmc/train.parquet",
			 "test":"https://huggingface.co/datasets/kraina/geolife/resolve/main/data/hmc/test.parquet"}
		)
	]
	
	def _info(self):
		return datasets.DatasetInfo(
			# This is the description that will appear on the datasets page.
			description=DESCRIPTION,
			homepage="http://research.microsoft.com/en-us/people/yuzheng/default.aspx",
			citation=CITATION,
			# This defines the different columns of the dataset and their types
			features=datasets.Features(
				{
					# "latitude": datasets.Value(dtype="float64"),
					# "longitude": datasets.Value(dtype="float64"),
					# "altitude": datasets.Value(dtype="float64"),
					# "time": datasets.Value(dtype="string"),
					# "timestamp": datasets.Value(dtype="int64"),
					# "mode": datasets.Value(dtype="string"),
					"trajectory_id": datasets.Value(dtype="string"),
					# "user_id": datasets.Value(dtype="string"),
					# These are the features of your dataset like images, labels ...
				}
			),
		)
	
	def _split_generators(self, dl_manager: datasets.download.DownloadManager):
		# files = _URLS[self.config.name]
		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