≡ Menu

First Look – Fuzzy Logix In-Database Analytics

Share

Fuzzy Logix is an advanced analytics software solution and services company headquartered in Charlotte North Carolina, with offices in Cupertino California, Richmond, Virginia and distribution channels throughout the world. Fuzzy Logix provides an in-database analytics solution, supporting both in-database model construction and in-database scoring.  They also offer in-GPU solutions, which allow users to access potentially hundreds of cores of graphics processing unit power. Fuzzy Logix offers consulting services related to the use of analytics; from helping companies adopt the use of analytics for the first time to custom model development for experienced teams.

DB Lytixs™, the Fuzzy Logix analytics engine, allows analytics to be deployed into various databases and data warehouses. The core of the product is a collection of 800+ algorithms written in C/C++ that can be embedded into a database kernel. These algorithms cover data mining, simulation, prediction, distributions and more.  Fuzzy Logix also offers FIN Lytix, a library of models that can be used by financial services companies to model risk and pricing for equities, fixed income and foreign exchange and also for portfolio optimization.

Both DB Lytix and FIN Lytix models run with simple SQL statements which execute deep in the database, allowing developers to use any development tool and any front end. SQL statements can be written that take these algorithms and integrate them with standard reports and processes, train models from historical data and return the model specification or take a model specification and apply it to the data in the database, creating real-time results. Both SAS (PROC SQL) and R can also be used to call the Fuzzy Logix in-database models.  The integration with the database ensures that the data being analyzed is up to date and that the execution of the models is done on the scalable, reliable infrastructure already in place for the database.  The product installs in 10 minutes and requires no additional hardware.

The in-database computation engine allows you to perform analytics within your database and eliminates the need to move data during model development and so reduces the computational infrastructure required.  It can also vastly shorten the time needed for model development; sometimes by a factor of 10X. . By moving analytic processing to the data, Fuzzy Logix produces enormous processing efficiencies and Fuzzy Logix believes that this profoundly changes the way companies leverage analytics.

Their Tanay GPU appliance and related software, GPU Lytix, allow companies to leverage the extreme calculation power of graphics processing units. The Tanay appliance comes with 2 or 4 GPU cards with each card having 448 cores (larger units are available). Using GPUs for processing usually means that users must learn CUDA; a parallel programming language for hardware, so Fuzzy engineered the models so they could be called via .NET or .DLL; making GPU processing more accessible.

DB Lytix™ and GPU Lytix contain a very rich collection of functions from simple statistical analysis to complex classification models such as multi-state support vector machines to pattern recognition and data mining functions such as linear/logistic regression. The available models include:

  • Advanced Mathematical Methods
  • Basic Statistics, Hypothesis Testing and Sampling
  • Data Mining and Predictive Analytics
  • Monte Carlo Simulation
  • Financial Analytics
  • Geo Spatial Analytics
  • Text Mining

The in-database models can be embedded into stored procedures and invoked using ODBC calls. The in-GPU models are invoked via .net or .dll.  The functions take advantage of the parallelism inherit in the each platform so they are fast and enable take advantage of the infrastructure already purchased, saving costs. For DB Lytix, the analytics run in the same processing thread as the SQL query.  This means the analytics are processed in the database and only the results are moved.

DB Lytix is integrated with a number of databases:

  • Sybase IQ
  • Netezza
  • Microsoft SQL Server
  • Informix
  • ParAccel
  • Aster Data
  • MySql

Fuzzy Logix also offers an in-database Scoring and Model Development Engine that supports aggregation and summarization of data, imputation of missing values, outlier identification, transformations and computation of predicted values and probabilities. There is also an Excel add-in to allow analysts who may not expert with writing SQL to set up and involve the algorithms on the tables or views in the database. The add-in builds the SQL dynamically, performs the analysis in-database and presents only the computed results in Excel.

Fuzzy Logix targets four distinct communities with business analytics. First the “quant” or analytic expert community can use the algorithms to rapidly create, iterate and deploy models. They can also use them as building blocks for more advanced algorithms. End users and analysts using dashboards and ad-hoc query tools are also potential users of the routines. Thanks to the integration with SQL, anyone writing SQL statements can use the algorithms and display the results alongside their standard data, integrate it with their graphs etc. Companies that would like to embed analytics into their products can work with Fuzzy Logix via OEM relationships to leverage their existing library. Finally Fuzzy Logix is working with companies who build databases and database appliances to pre-integrate their algorithms and so make it easier for users of those systems to deploy data mining and predictive analytics.

Fuzzy Logix has number of applications based on its algorithms including ones for customer acquisition and retention, churn reduction, scoring, risk management, forecasting, web analytics and optimization.

Fuzzy Logix  is one of the vendors listed in our Decision Management Systems Platform Technologies report.

Share

Comments on this entry are closed.