New and improved forecasting engines
At the heart of Release 11 are the calculation methods, known as engines. Three are new and three have been improved. At the same time, all engines now have consistent names:
- mPX Spline (formerly mPX Regression): improved
- mPX Neural (formerly ANN): completely re-implemented
- mPX Compare (formerly Similar Day Comparison): improved
- mPX Neighbor (formerly Nearest Neighbor): improved
- mPX Horizon (N-BEATSx): new
- mPX Forest (LightGBM): new
mPX Horizon is built for target variables with pronounced daily, weekly or seasonal cycles. The engine combines these patterns with trends, automatically selects the most relevant influencing factors and also captures complex intraday patterns.
mPX Forest belongs to the family of decision tree methods and is particularly well suited to complex consumption profiles. It is robust against outliers and noisy data and handles categorical factors such as weekdays and public holidays especially well.
mPX Neural replaces the previous artificial neural network. A new optimization method improves training, and the engine works well with raw data, without additional feature engineering.
mPX Spline takes the proven regression method a step further: faster, with an almost unlimited number of influencing factors and reliable even with little training data.
Faster, more stable, more secure
A lot has also changed under the hood. The most important improvements in the mP Xnergy Client:
- New mP Optimizer that automatically determines the best possible model
- Faster loading of projects and large data volumes thanks to optimized interfaces to MS SQL Server, Oracle and PostgreSQL
- Lower storage requirements through compression and distributed data storage
- Extended automation and more functions in the WebAPI
- Improved logging and better error handling
- Stronger protection against attacks in line with current IT security requirements, with high availability on request
Built for the cloud
mP Xnergy is developed as a cloud-native application, designed from the ground up to run in a cloud environment such as Microsoft Azure. Release 11 can still be run on-premises, but some functions perform faster and more efficiently in the cloud.
Still running mP Xnergy on your own servers? We’d be happy to show you how the move to Software as a Service works and what you gain from it.
Outlook: what’s next
Development continues. The following is planned for upcoming versions:
- Probabilistic forecasts with P10, P50 and P90 for well-founded risk management
- Script management with versioning for metaScript Master and mP Workflow
- New web client for easier operation
- mPX extension for Python, so you can use your own Python models directly in mP Xnergy
Get in touch
Do you have questions about Release 11 or want to know what mP Xnergy in the Azure Cloud can do for you? Contact us, we’ll be happy to advise you.
