Introducing Forecasted Metrics:
Unleashing the power of AI

We are excited to announce the beta release of our latest feature called Forecasted Metrics! It is a pioneering feature which adds AI capabilities to the Blockbax Platform. A forecasted metric allows you to forecast future values of an existing metric. Your existing data (including other metrics and properties) is utilized to train a forecasting model which is capable of forecasting values in real-time.

No-code approach with AutoML

At Blockbax we aim to abstract and automate as much as possible to enable faster delivery. This makes features more accessible to users without technical expertise and automates the mundane tasks for technical experts. In machine learning (ML) this domain is called AutoML and this feature is built upon the AutoML methods and processes.

In order to understand the complexity of creating a forecasting model, an ML-expert would need to:

1. Pre-process and clean the data,
2. Select and construct appropriate features,
3. Select an appropriate model family (ML technique),
4. Optimize model hyper parameters,
5. Design the topology of neural networks (if deep learning is used) and
6. Post-process the results.

Typically an ML-expert would do several iterations to get to an acceptable model.

Forecasting with Blockbax takes away much of the underlying technicalities and with its userfriendliness it is offered in a way that it can be used by everyone:

• Simple configuration
• Flexible forecast periods ranging from seconds to weeks
• Leveraging existing data from properties and metrics, such as adding correlated time series like solar irradiance for temperature forecasts, or using static property data like window orientation to improve indoor temperature predictions.
• Scalability to train on large amounts of historical data.
• Automatic data cleansing / preparation (takes most time normally): Unlike many machine learning models that require uniformly sampled data, this feature is forgiving when faced with datasets that are not rich, flaky, highly irregular and with missing values. By minimizing the need for tampering with raw data, Blockbax enables users to scale their models effortlessly and obtain reliable insights even with complex and diverse datasets.

Advantages over similar tools

• Coherent experience within its streaming processes which enables real-time forecasts.
• AI model over linear regression, which can be more adept at capturing complex patterns.
• Accurate in complex scenarios.
• Continuous performance assessment and model retraining.

Outperforming similar tools

Constantly striving to benchmark and improve, we recently evaluated our forecasting solution against NIXTLA’s StatsForecast and Amazon Forecast, using the Walmart M5 dataset and Blockbax Climate Monitoring data. Our solution exhibited lower errors and improved scalability compared to StatsForecast. For detailed benchmark analysis, you can refer to this news post . We’re continuously refining our methods to ensure accuracy and welcome further comparisons.

Use cases

Forecasted metrics can enable users in decision making based on the uncovered underlying patterns, trends, and seasonality of data. The generalized implementation, makes it versatile for various industries, for example:

• Energy & Utilities
• Agriculture
• Healthcare
• Transportation & Fleet
• Supply Chain & Logistics
• Finance & Stock Market

A few useful examples are:

• Temperature forecast of rooms to create more efficient heating and cooling.
• Grid voltage anomaly detection to maintain grid stability and prevent outages.
• Forecast soil humidity level for more efficient irrigation planning.
• Forecast supply and demand.

Conclusion

We have introduced the Forecasted Metrics. With this new feature, you can make predictions to help guide your decision-making processes. It relies on neural network deep learning models to interpret historical data and forecast future outcomes. This integration is designed to enhance your workflow by proactively identifying possible challenges before they arise. This feature is currently accessible in private beta as we fine-tune its performance for a wide range of applications, and benchmark it against more tools.

If you are interested to know more about this feature or have any questions or notes, please don’t hesitate to contact us

What next?

We did a benchmark and we are not inferior, in fact we perform a little better. More on that in our next blog!

Cheers,
The Blockbax Team.