VIA Analytics

VITRIA VIA Analytics CoreThe diversity of IoT use cases demands a range of analytics capability, spanning from traditional descriptive analytics to more advanced predictive and prescriptive analytics. VIA supports an extensive set of analytic techniques that meet the demands of IoT. VIA Analytics includes all the key types of analytics – real-time, historical, predictive, and prescriptive – needed for IoT. It executes fast analytics in real-time to provide the context and insight needed for the fast decision-making required in IoT.

 

Highlights of VIA’s Analytic capabilities include:

  • Analytics over real-time streams of data and over historical (batch) data.
  • Large portfolio of descriptive, predictive and prescriptive analytic techniques.
  • Time-series analysis provides insight into the behavior of IoT networks over time.
  • Analytic data quality even in the presence of lost, delayed, and out-of-order events.
  • Complete Machine Learning environment for building predictive and prescriptive models.
  • Rapid operationalization of predictive and predictive models
  • Open Machine Learning environment that supports the importation and operationalization of 3rd party Machine Learning models via standards.
  • Rapid innovation through incremental refinement of Machine Learning models through rapid iterations of the discover → model → operationalize → monitor modeling cycle.
  • Visual modeling environment to enable the development of IoT analytic solutions, in hours not months, and with little or no coding.
  • Production-grade analytics executing at the scale and speed of even the largest IoT Networks

 

Real-Time and Historical Analytics

VITRA VIA Real-Time and Historical AnalyticsVIA supports real-time continuous processing of all analytics – including descriptive, predictive, and prescriptive analytics – for the many IoT use cases that are time-sensitive. Real-time descriptive analytics describes the world as it is right now and provides contextual awareness and situational intelligence; real-time predictive analytics predicts what will happen next; while real-time prescriptive analytics prescribes the next best actions to optimize business outcomes.

VIA’s real-time analytics supports the scale and speed of the most demanding IOT use cases – often involving millions of events per second with fast, subsecond processing latency.

The same analytical techniques available for real-time analytics are also available for historical analytics and batch processing. Historical analytics provides the historical context for interpreting real-time analytics, baselines for anomaly detection, and input for machine learning.

A major strength of the VIA Analytics Platform is its ability to build equivalent analytic pipelines to run either continuously over streaming data producing true real-time analytics, or in batch over large historical data sets.

 

Descriptive Analytics

VITRA VIA Descriptive AnalyticsDescriptive Analytics provide a description of the world as it is right now (real-time) or as it was in the past (historical). VIA’s descriptive analytics includes KPIs and baselines, statistical summaries, multidimensional analysis, pattern matching, anomaly detection, trend analysis, and behavioral analytics. Descriptive analytics can be performed either continuously in real-time over streaming data or periodically over large batches of data.

 
A more comprehensive list of VIA’s descriptive analytics capabilities can found below:

  • Correlation
  • KPIs
  • Multidimensional Analysis
  • Summary Statistics
  • Anomaly Detection
  • Geospatial
  • Pattern Matching
  • Trending
  • Time-series Analytics
  • Population Analytics
  • Activity Analytics
  • Behavioral Analytics
  • Track and Trace
  • Link Analysis
  • Hypothesis Testing
  • Root Cause Analysis

 

Predictive and Prescriptive Analytics

VITRA VIA Predictive and Prescriptive AnalyticsVIA’s Predictive Analytics supports regression, classification, and clustering using hundreds of predictive techniques based on machine learning algorithms.

VIA’s Prescriptive Analytics leverages rules-based and machine learned prescriptions to recommend the next best action based on the current situation and latest predictions. Hundreds of prescriptive techniques are available.

VIA can score predictive and prescriptive models in real-time (streaming) or batch mode, and features elastic scaling over big and fast data.

 

Machine Learning

VITRA VIA Machine LearningVIA’s Machine Learning provides a rich and flexible environment for continuous learning and refinement. Machine learning is executed over historical data in the VIA Open IoT Data Lake to produce predictive and prescriptive models. VIA’s Machine Learning capabilities include:
 
 

  • Supervised and unsupervised learning
  • A large repertoire of classification, regression, and clustering algorithms
  • Visual design of analytic pipelines for model building and iterative refinement.

VIA’s supports a wide range of machine learning algorithms for building predictive and prescriptive models. Some of the most popular Machine Learning Algorithms supported include:

  • Clustering
  • Neural Network
  • Regression (linear)
  • Logistic regression
  • Decision Tree
  • Support Vector Machine
  • Random Forest
  • Association Rules
  • Naïve Bayes Classification
  • Time Series (ARIMA, …)
  • Exponential Smoothing
  • k-Nearest Neighbors
  • Scorecard Model
  • Rule Set Model
  • Plus many more …

 

Rapid Operationalization of Machine-Learned Models for Prediction and Prescriptive Actions

VITRA VIA Rapid Operationalization of Machine-Learned Models for Prediction and Prescriptive ActionsRapid operationalization of machine learned models into an analytics pipeline for prediction and prescriptive next actions is critical to fast time to value in IoT. In VIA, new or refined Machine-Learned models can be put into operations in minutes – not days – utilizing the following capabilities:
 
 

  • An open environment supporting the direct importation of Predictive & Prescriptive Models via standards, e.g., PMML, and popular tools such as R
  • Fast deployment of natively built and imported models
  • Warm deployment of new and refined models through simple administrative controls