Long time taken for the Multivariate Anomaly Detection

[waqarkoc]

• Orion version:
• Python version:
• Operating System: Google Colab

Description

I have 5 columns dataset https://github.com/microsoft/fabric-samples/blob/main/docs-samples/real-time-intelligence/demo_stocks_change.csv . It takes alot of time for detecting anomalies

I have following code to run Anomalies

from orion.data import load_signal

signal_path = '/content/demo_stocks_change.csv'
import pandas as pd
df = pd.read_csv(signal_path)
df.rename(columns={'Date': 'timestamp'}, inplace=True)
df['timestamp'] = pd.to_datetime(df['timestamp'])
from orion import Orion

hyperparameters = {
    "mlstars.custom.timeseries_preprocessing.time_segments_aggregate#1": {
        'interval': 1
    },
    'orion.primitives.aer.AER#1': {
        'epochs': 5,
        'verbose': True
    }
}

orion = Orion(
    pipeline='aer',
    hyperparameters=hyperparameters
)

orion.fit(df)

It takes more than 1 hour but I can not see any epoch progress in the output

[sarahmish]

Thank you @waqarkoc for using Orion!

It might be a large dataset, can you try to see if you limit the length of df to somewhere around a thousand entries, do you see the progress bar?