Network state prediction on SAP HANA

Incelligent, Wings ICT Solutions’ telecommunications spin – off, has released a network state predictor on SAP HANA

Incelligent on SAP hanaIncelligent participates in SAP’s StartupFocusProgram, an initiative supporting startups that provide solutions on predictive analytics using SAP Hana. Incelligent’s implementation focuses on accurate network state predictions leveraged by heterogeneous, seemingly uncorrelated data sources.

It addresses network operators’ challenge to:

(a) allocate network resources so as to offer the highest possible QoS to all of his clients and/ or to be more energy efficient,

(b) run targeted promotions of free or discounted services to selected time and space, in order to maximize conversion to subscription (i.e. predict best candidates for the promotion) while minimizing the impact to existing clientele (i.e. predict network underutilization).

Features

The SAP Hana ready solution provides the following features:

  • Builds the pattern of the network state (load, performance, QoS, etc.).
  • Leverages highly heterogeneous data including weather, area, day and time access, user and app data, etc.
  • Predicts based on unsupervised machine learning techniques including parameter-less growing and traditional self-organizing maps
  • Provides User-friendly visualized diagrams of the network state vs time
  • Supports access from any device based on HTML 5
  • Uses SAP HANA as primary DB and warehouse, dynamic creation of tables and R server for maximizing the speed of processing
  • Supports power users - sets variables for predicting the network load, builds new knowledge, validates the tool, and sets requests for network load prediction
  • Supports business users - sets requests for network load predictions

Business Value

  • Management automation and energy savings lowering operational expenditure by 30-40%
  • Reduce capital expenditures 10% - 20% by avoiding worst-case oriented planning
  • Increase revenue by leveraging predictions to offer personalized services very efficiently
  • Achieves prediction accuracy at the level of 85-95% and ultra-low deviations, typically for a time – windows of 7-15 days.
  • Enhances agility, speed, reliability, and proactivity
  • No expertise required - Easily configurable and extensible

Demo

The following demo video illustrates the use of the tool as implemented over the SAP HANA platform. The application in this case is the prediction of network load based on time, space and weather.