doubt.datasets.facebook_comments

  1"""Facebook comments data set.
  2
  3This data set is from the UCI data set archive, with the description being the original
  4description verbatim. Some feature names may have been altered, based on the
  5description.
  6"""
  7
  8import io
  9import zipfile
 10
 11import pandas as pd
 12
 13from .dataset import BASE_DATASET_DESCRIPTION, BaseDataset
 14
 15
 16class FacebookComments(BaseDataset):
 17    __doc__ = f"""
 18    Instances in this dataset contain features extracted from Facebook posts. The task
 19    associated with the data is to predict how many comments the post will receive.
 20
 21    {BASE_DATASET_DESCRIPTION}
 22
 23    Features:
 24        page_popularity (int):
 25            Defines the popularity of support for the source of the document
 26        page_checkins (int):
 27            Describes how many individuals so far visited this place. This feature is
 28            only associated with places; e.g., some institution, place, theater, etc.
 29        page_talking_about (int):
 30            Defines the daily interest of individuals towards source of the
 31            document/post. The people who actually come back to the page, after liking
 32            the page. This include activities such as comments, likes to a post, shares
 33            etc., by visitors to the page
 34        page_category (int):
 35            Defines the category of the source of the document; e.g., place,
 36            institution, branch etc.
 37        agg[n] for n=0..24 (float):
 38            These features are aggreagted by page, by calculating min, max, average,
 39            median and standard deviation of essential features
 40        cc1 (int):
 41            The total number of comments before selected base date/time
 42        cc2 (int):
 43            The number of comments in the last 24 hours, relative to base date/time
 44        cc3 (int):
 45            The number of comments in the last 48 to last 24 hours relative to base
 46            date/time
 47        cc4 (int):
 48            The number of comments in the first 24 hours after the publication of post
 49            but before base date/time
 50        cc5 (int):
 51            The difference between cc2 and cc3
 52        base_time (int):
 53            Selected time in order to simulate the scenario, ranges from 0 to 71
 54        post_length (int):
 55            Character count in the post
 56        post_share_count (int):
 57            This feature counts the number of shares of the post, how many people had
 58            shared this post onto their timeline
 59        post_promotion_status (int):
 60            Binary feature. To reach more people with posts in News Feed, individuals
 61            can promote their post and this feature indicates whether the post is
 62            promoted or not
 63        h_local (int):
 64            This describes the hours for which we have received the target
 65            variable/comments. Ranges from 0 to 23
 66        day_published[n] for n=0..6 (int):
 67            Binary feature. This represents the day (Sunday-Saturday) on which the post
 68            was published
 69        day[n] for n=0..6 (int):
 70            Binary feature. This represents the day (Sunday-Saturday) on selected base
 71            date/time
 72
 73    Targets:
 74        ncomments (int):
 75            The number of comments in the next `h_local` hours
 76
 77    Source:
 78        https://archive.ics.uci.edu/ml/datasets/Facebook+Comment+Volume+Dataset
 79
 80    Examples:
 81        Load in the data set::
 82
 83            >>> dataset = FacebookComments()
 84            >>> dataset.shape
 85            (199030, 54)
 86
 87        Split the data set into features and targets, as NumPy arrays::
 88
 89            >>> X, y = dataset.split()
 90            >>> X.shape, y.shape
 91            ((199030, 54), (199030,))
 92
 93        Perform a train/test split, also outputting NumPy arrays::
 94
 95            >>> train_test_split = dataset.split(test_size=0.2, random_seed=42)
 96            >>> X_train, X_test, y_train, y_test = train_test_split
 97            >>> X_train.shape, y_train.shape, X_test.shape, y_test.shape
 98            ((159211, 54), (159211,), (39819, 54), (39819,))
 99
100        Output the underlying Pandas DataFrame::
101
102            >>> df = dataset.to_pandas()
103            >>> type(df)
104            <class 'pandas.core.frame.DataFrame'>
105    """
106
107    _url = (
108        "https://archive.ics.uci.edu/ml/machine-learning-databases/" "00363/Dataset.zip"
109    )
110
111    _features = range(54)
112    _targets = [53]
113
114    def _prep_data(self, data: bytes) -> pd.DataFrame:
115        """Prepare the data set.
116
117        Args:
118            data (bytes): The raw data
119
120        Returns:
121            Pandas dataframe: The prepared data
122        """
123
124        # Convert the bytes into a file-like object
125        buffer = io.BytesIO(data)
126
127        # Unzip the file and pull out the csv file
128        with zipfile.ZipFile(buffer, "r") as zip_file:
129            csv = zip_file.read("Dataset/Training/Features_Variant_5.csv")
130
131        # Convert the string into a file-like object
132        csv_file = io.BytesIO(csv)
133
134        # Name the columns
135        cols = (
136            ["page_popularity", "page_checkins", "page_talking_about", "page_category"]
137            + [f"agg{n}" for n in range(25)]
138            + [
139                "cc1",
140                "cc2",
141                "cc3",
142                "cc4",
143                "cc5",
144                "base_time",
145                "post_length",
146                "post_share_count",
147                "post_promotion_status",
148                "h_local",
149            ]
150            + [f"day_published{n}" for n in range(7)]
151            + [f"day{n}" for n in range(7)]
152            + ["ncomments"]
153        )
154
155        # Read the file-like object into a dataframe
156        df = pd.read_csv(csv_file, header=None, names=cols)
157        return df