PredictHQ's Beam

The automated correlation engine to quickly and accurately reveal the events that drive demand for your business.

Beam turns 72 days of work into 54 seconds

  • 1

    Impute values + check quality

    0.6 seconds vs. 2.5 days
  • 2

    Time-series decomposition

    48 seconds vs. 45 days
  • 3

    Anomaly
    detection

    2 seconds vs. 7 days
  • 4

    False-positive
    detection

    1.5 seconds vs. 7 days
  • 5

    Correlate
    demand

    2 seconds vs. 10 days

The Input

Combine unstructured event data + transactional data

We convert unstructured, dynamic event data into a workable dataset, allowing Data Science teams to use our intelligent event data in machine learning models. We then combine and model our event data with your transactional demand data. This becomes the input for Beam.

Step 1: Impute values + check quality

Beam automatically detects missing values from your transactional data and imputes values using our advanced time series re-construction technologies. We then systematically check data quality with stationary attributes and rolling standard deviation.

Learn more

Step 2: Time-series decomposition

We decompose your time-series transactional data using IteSSA nonparametric machine learning algorithms to automatically detect weekly and monthly cyclicities, seasonal patterns, and long term and short term trends.

Learn more

Step 3 + 4: Detection

Anomaly Detection uses the patterns identified in step two, Beam automatically detects the positive anomalies in the transactional data based on the re-constructed demand patterns.

False-positive detection algorithms detect the positive anomalies which are not generated by attendance-based events and then adjust the incremental demand estimation.

Learn more

Step 5: Correlate demand

Beam uses the False Discovery Rate to calculate the strength of the incremental demand. First it segments all detected incremental demand into strong, medium, weak, and no impact. It then combines the detected incremental demand with attendance-based events impact and holiday events.

Learn more

The Output

Export the data and seamlessly integrate it into your prediction models

Beam has successfully correlated your transactional demand data with PredictHQ’s event data. You can now download the full results and use the results to train your demand forecasting models so you can predict future demand easily and accurately.

img-correlation-graph

Reveal signals, remove noise. Fast.

Contact us to get access to our automated correlation engine today.

Get Started

Contact us now to find out the best solutions for your business. We'll get back to you within 1 working day.

Trusted By

  • Accenture
  • Wyndham
  • Booking.com
  • Amadeus
  • Qantas
  • Domino's

For other questions

  • * fields are required

PredictHQ Preparedness Kit - A Guide to Navigate COVID-19 Impact and Plan for Recovery