office@audascience.com +972-77-8011500

Spark & R Challenges and Opportunities

Legacy tools that were once only accessible to large corporations, are constantly being replaced by cost-effective and accessible open source platforms and programming languages, created and continuously improved by developer communities. As the Open Source solution for analytics and, more specifically, for predictive analytics and machine learning, Spark is becoming a popular and effective framework for resolving multifaceted commercial and statistical challenges.

New trends in Machine Learning

Organizations are late to adopt innovations in machine learning - in spite of dramatic changes in business strategies and practices.

In recent years, new abilities of storing, processing and analyzing huge amounts of diverse data, with lower costs and reduced convoluted demands on management, are shaking-up the classic data warehouse realm. Modern Visualization tools are taking over the complex, rigid and time-consuming legacy business intelligence processes. These new tools are so user-friendly they succeed in making the analysis process accessible to business level operators.

Technology Edge

AuDaScience BRAINs™ [Big-data Recommendations for Actionable Insights] applications are based on advanced technologies in the area of big data analytics.

The use of the Spark engine

statistical computing language‎

AuDaScience BRAINs applications operate the Spark engine for automatically running thousands of parallel processes using the Spark cluster, enabling high scalability and superior performance. The automatic data management, modeling, machine learning and batch and real-time scoring processes run using Spark SQL, Spark ML and Spark Streaming with embedded R components. In this manner, the Spark engine performs all data management tasks, transformations, calculations and modeling in-memory instead of performing them in a relational DB.

AuDaScience Multi-Segment-Modeling Technology

AuDaScience patent-pending Multi-Segment-Modeling technology is the "secret sauce" which allows designing and deploying propensity models on customer segments, rather than on whole populations. By using this technology, AuDaScience customers enjoy model lifts of up to 50% higher, in comparison to manually-developed models using any data mining tool. This leads to higher response rates in outbound and inbound campaigns, augmented revenues from targeted campaigns and visible ROI after only one or two campaigns.

In-memory processing

AuDaScience BRAINs applications operate the Spark engine's in-memory analytical processing technology. All model development and deployment processes are done in-memory, without writing to any system tables. This allows significant cost savings in disk space and processing time compared to traditional modeling methods that use common machine algorithms tools and create a multitude of large, temporary tables throughout the process.

Connectivity to Common Big Databases

The Spark engine embedded in AuDaScience BRAINs application can read data files from any HDFS resources and can connect to most common databases.

AuDaScience BRAINs applications can also work with common relational data bases, using a set of connectors (for Oracle, Exadata, Teradata, SQL Server, Netezza, Vertica, MySql). There is no transfer of huge data sets between servers and environments and thereby, optimal usage of the DB analytical processing power is achieved.

Cloud Integration

The AuDaScience BRAINs applications can be implemented and run on-site, integrated with the company's IT environment, (DWH, big data databases and campaign management tools) or on the cloud (AWS Amazon).

Machine Learning and Marketing Optimization for Commercial Banking

How to optimize marketing campaigns so that at any time, each customer is offered the services and products with the highest probability for conversion and highest profitability to the bank.