Source code for torch_timeseries.dataset.M4

# https://github.com/Mcompetitions/M4-methods/tree/master/Dataset


import os
import resource
from ..core.dataset.dataset import Dataset, TimeSeriesDataset
from typing import Any, Callable, List, Optional
import torch
from torchvision.datasets.utils import (
    download_url,
    download_and_extract_archive,
    check_integrity,
)
import pandas as pd
import numpy as np
import torch.utils.data


[docs]class M4(TimeSeriesDataset): """ The M4 forecasting competition, the continuation of the previous three ones organized by Spyros Makridakis (https://en.wikipedia.org/wiki/Makridakis_Competitions). collected from https://github.com/Mcompetitions/M4-methods/tree/master,we concate train,test files as the dataset. Parameters: category (str): The category of the dataset to load. Can be one of ['Daily', 'Hourly', 'Monthly', 'Quarterly', 'Weekly', 'Yearly']. Default is 'Daily'. root (str): The root directory to store the dataset. Default is './data'. Attributes: dir (str): The directory where the dataset is stored. dates (pd.DataFrame): The dates associated with the dataset. df (pd.DataFrame): The combined training and test data. data (np.ndarray): The time series data. Methods: download(): Download the dataset from the M4 repository. _load(): Load the dataset from the local files. """ name: str = "M4" def __init__(self, root: str = "./data", category="Daily"): self.category = category self.root = root self.dir = os.path.join(root, self.name) os.makedirs(self.dir, exist_ok=True) self.download() self._process() self._load() self.dates: Optional[pd.DataFrame] def _download_category(self, category): download_url( f"https://raw.githubusercontent.com/Mcompetitions/M4-methods/master/Dataset/Train/{category}-train.csv", self.dir, filename=f"{category}-train.csv", ) download_url( f"https://raw.githubusercontent.com/Mcompetitions/M4-methods/master/Dataset/Test/{category}-test.csv", self.dir, filename=f"{category}-test.csv", ) def download(self): self._download_category(self.category) def _load(self) -> np.ndarray: self.train_file_path = os.path.join(self.dir, f"{self.category}-train.csv") self.test_file_path = os.path.join(self.dir, f"{self.category}-test.csv") self.tarin_df = pd.read_csv(self.train_file_path) self.test_df = pd.read_csv(self.test_file_path) self.df = pd.concat([self.tarin_df, self.test_df]).dropna(axis=1) self.data = self.df.iloc[:, 1:].to_numpy() return self.data