TECHNOLOGY
Vital Analytix

VITAL ANALYTIX

Flexible and Modular

  • Architecture: Three-tier cloud-enabled architecture with Application, Web, and DB Servers. Built on J2EE design patterns with business logic and data access separated
  • Works with all industry standard operating systems, application servers and databases
  • Mobility: Web-based access from desktop, mobile or any web-enabled device

Analytical

  • Multi-dimensional Analysis enabled by creation of smart data marts and OLAP Cubes
  • Master Data Management enabled via front-end access to users for editing, updating and deleting
  • MDX engine enables business users to perform multi-dimensional analysis using a Slice and Dice interface

Scalable and Reliable

  • Ability to handle large data sets without compromising on performance
  • Easily manage large number of concurrent users
  • Role-based data access and extensive audit trails
Workflow Engine

Workflow Engine

Core Engine

  • Based on industry standard JBPM framework
  • Possible to customize business process rules
  • Integration with email engine for alerts
  • Includes Data Entry screens and front-end uploads
  • Extensive logs for audit and other purposes

Application

  • Integrate into your Business Process: Capture tactical knowledge inputs which are not currently being captured by business processes or transaction processors
  • Case Management: Intelligence for decision making such as approve, reject, escalate, forward, etc.

Use Case examples

  • Budgeting and Planning
  • Case Manager for Abnormal Transaction behavior such as Anti-Money Laundering and Fraud Analytics
  • Data upload for Master Data Maintenance
Big Data

Big Data

VITAL CLUSTER

  • A load balanced OLAP architecture to scale up processing capability
  • In-memory computing and custom pagination to project large datasets to the end user

IN-MEMORY CACHE

  • Fast Cache grid implementation using Redis
  • Stores large datasets for the MDX queries
  • Significant improvement in performance when data is preloaded as part of the BoD/EoD ETL run

NoSQL based OLAP

  • Move past the OLAP engine constraint of querying RDBMSes by implementing NoSQL querying capability
  • MDX to NoSQL translation will bring to the table the Big Data framework benefits
  • MongoDB document stores and HBase based solutions
  • Hadoop/Spark frameworks used to aggregate large data-sets, to service end user queries
Visualization

Visualization

Dashboard Designer

  • Design-it-yourself: Create custom dashboards with 100+ customizable charts
  • Interactive: Dig deeper using dimension filters and drill-down capabilities
  • Favorites: Add important KPIs to the page for a running indicator

Slice and Dice

  • Create data pivots that include sub-totals, filters, conditional formatting, data sorting
  • Paginate datasets to simplify navigation
  • Save and Schedule views to reach multiple users
  • Download to Microsoft Excel, CSV, PDF

Decision support and Reporting

  • Access on the desktop browser, mobile device and through plug-ins for Microsoft Excel and Outlook
  • Decision support capability including alerts and exception reports
  • Report designer functionality

Vital Portal

  • An add-on module, that provides single view to multiple indicators of the business
  • Combines and harmonizes data from diverse and multiple data sources to present complex KPIs
  • Custom-designed user interface
  • Drill-down capability to reach the atomic data level
ETL

ETL

Connects and Reads from diverse sources

  • Extract from disparate sources such as structured databases, flat files, XML, unstructured data and web/FTP servers
  • Standard and custom validation and cleansing capabilities including entity resolution
  • Data storage techniques include Normalization, Encryption and Decryption

Extensive Transformational Capabilities

  • Logical, statistical, mathematical functions for complex transformational capabilities
  • Generate new fields or values based on aggregate data, or defined functions
  • Store calculated values for faster retrieval
  • Combine data from multiple sources to create new data sets

Smart Execution

  • Batch job execution for parallel ETLs
  • Scheduled as well as on-demand transformations with detailed logs and post-transformation alerts
  • Automated mails to specific lists
  • Auto-delete after a process
  • GUI-based interface for ease of design
Data Modeling

Data Modeling

Feature Engineering

  • Deep domain expertise results in creation of relevant derived attributes
  • Enhanced variable selection using statistical modelling techniques
  • Model selection, verification & validation for balanced fitment

Integrated Solution

  • Results of the data modelling is integrated with the reporting/visualization framework
  • Ease of interpretation and decisioning
  • Automated email of training and prediction results with complete details including error curves

Parameterization & Scalability

  • Facilitates user defined parameters for consideration in the analytical framework
  • Data quality monitoring
  • Ability to support large volumes of data on both traditional and big data stacks
Algorithm Description Use Case
Principal Component Analysis Combines multiple explanatory variables into a representative few for better understanding underlying phenomena, i.e. Exploratory analysis of data sets
  • Product pricing analytics
  • Credit scoring
  • Bad debt prediction
  • Customer Churn
  • Revenue/spend forecasting
  • Dunning efficiency
Correlation Analysis Identify linear inter-relationships among variables during exploratory analysis
  • Product Pricing Analytics
Association Rule Mining Finds close relationships between two sets of occurrences/events (identify market-basket)
  • Cross-sell
  • Up-Selling products
K-means clustering Segmentation based on transaction behavior or similarity of customer attributes
  • Product Pricing Analytics
  • Portfolio Insights
Logistic & Linear regression A generalized linear model for classification of events
  • Product pricing analytics
  • Credit scoring
  • Bad Debt
  • Churn
Classification and Regression Trees A classifier which builds a decision tree based on historic examples, new cases are predicted using the decision tree
  • Customer response to dunning actions
  • Churn Prediction
Neural Networks A set of simple Artificial Neurons organized into a network which mimics the biological neural system to classify events
  • Customer response to dunning actions
  • Churn Prediction
Support Vector Machine Highly optimized classifier which transforms feature space dimensions by using different types of kernel functions.
  • Customer payment behavior prediction
Stochastic Gradient Descent Highly accurate tree based classifier with optimizer searching for minimum out-of-Bag training error
  • Credit scoring and Early warning
Random Forest Ensemble of multiple decision trees organized to minimize error rates and increase accuracy by combining different types of classifiers
  • Predicting Customer behaviour basis 360 view of transaction/attributes