Time Series Processing

Endpoint: POST /timeseries

Request Parameters

File Upload

• file (required): The file to be processed. Supported formats include CSV, Excel, etc.

Form Parameters

• datetime_column (optional): The column containing datetime values. If not provided, the API will attempt to auto-detect datetime columns.
• value_columns (optional): A list of columns containing numeric values to be processed.
• resample_frequency (optional): The frequency for resampling (e.g., D for daily, H for hourly). If provided, resampling is applied.
• resample_method (optional, default: mean): The method for resampling. Supported values:
  • mean: Resample using the mean.
  • sum: Resample using the sum.
  • ffill: Forward fill missing values.
  • none: No resampling.
• timezone_handling (optional, default: keep): How to handle timezones. Supported values:
  • keep: Keep the original timezone.
  • convert_to_utc: Convert datetime values to UTC.
  • strip: Remove timezone information.
• time_features (optional): A comma-separated list of time-based features to extract. Supported values:
  • hour: Extract the hour.
  • day_of_week: Extract the day of the week.
  • month: Extract the month.
  • quarter: Extract the quarter.
  • day_of_year: Extract the day of the year.
• lag_periods (optional, default: 1,7,30): A comma-separated list of lag periods to create.
• rolling_windows (optional, default: 3,7): A comma-separated list of rolling window sizes.
• rolling_aggregations (optional, default: mean,std): A comma-separated list of aggregation methods for rolling windows. Supported values:
  • mean: Rolling mean.
  • std: Rolling standard deviation.
  • sum: Rolling sum.
  • max: Rolling maximum.
  • min: Rolling minimum.
• window_size (optional, default: 30): The size of the rolling window.
• forecast_horizon (optional, default: 1): The forecast horizon for creating future values.

Processing Logic

1. Read the uploaded file: The file is converted into a Pandas DataFrame.
2. Auto-detect datetime columns: If datetime_column is not provided, the API attempts to detect datetime columns.
3. Handle timezones: Applies the specified timezone handling (keep, convert_to_utc, or strip).
4. Resample data: If resample_frequency is provided, the data is resampled using the specified method (mean, sum, ffill, or none).
5. Extract time features: If time_features is provided, extracts the specified time-based features.
6. Create lag features: Creates lag features for the specified lag_periods.
7. Apply rolling windows: Applies rolling windows with the specified sizes and aggregation methods.
8. Create forecast horizon: Adds a forecast column shifted by the specified forecast_horizon.
9. Return the modified file: The processed file is provided for download.

Example Request

curl -X POST "http://localhost:8000/timeseries" \
-F "file=@data.csv" \
-F "datetime_column=date" \
-F "value_columns=value" \
-F "resample_frequency=D" \
-F "resample_method=mean" \
-F "timezone_handling=convert_to_utc" \
-F "time_features=hour,day_of_week" \
-F "lag_periods=1,7,30" \
-F "rolling_windows=3,7" \
-F "rolling_aggregations=mean,std" \
-F "window_size=30" \
-F "forecast_horizon=1"