11Ants Analytics Introduces Revolutionary Method of Building Ensembles at Predictive Analytics World, London
HAMILTON, New Zealand, 15 November 2010 - 11Ants Analytics today introduced automated ensemble building, a new feature in 11Ants Model Builder - a powerful data mining and predictive analytics tool. The new feature allows users to quickly build and evaluate thousands of ensemble permutations. Ensembles are recognized to increase the predictive power of models, and are a valuable tool in any data mining expert's arsenal. The announcement was made at its U.K launch at Predictive Analytics World conference in London.
"The challenge has been that construction of the best performing ensembles has required a good working knowledge and been an extremely time consuming task - with the final result usually representing a compromise between time available and predictive accuracy achieved," says Tom Fuyala, Director, Business Development, 11Ants Analytics. "With the automated ensemble building feature in 11Ants Model Builder this problem has been eliminated."
Through its revolutionary approach, 11Ants Model Builder enables anyone who can use Excel to mine their own data and build powerful predictive models; with the power and automation to benefit the most advanced analytic researchers and practitioners. This assists in providing more accurate insights to business, government and science challenges from recommendation systems to fault diagnosis, and scientific research to fraud detection.
"Users are impressed by the approaches 11Ants Model Builder offers, not least the ease of use. Users particularly like that they can build models quickly within Excel, without the set up involved in previous generation tools," says Mr. Fuyala. "The addition of this feature makes ensembles accessible to people who don't even know exactly what an ensemble is, let alone understand the inner workings of one. The true value is that users are empowered to increase predictive accuracy without having to be schooled on the details of the latest research - this is what is important."
11Ants Model Builder opens whole new possibilities for even the first time data miner to unlock useful information from data. The software's benefits are not limited to non-technical users; experienced data miners are freed up from traditionally time consuming and repetitive tasks with features such as automated algorithm selection, automated tuning parameter optimization, automated variable selection, automated model evaluation and now automated ensemble building.
11Ants Model Builder simplicity is due to its embedded HyperLearn technology, which does the complex work for the user, and is based on a library of 11 machine learning algorithms including decision tree (similar to CART), Gaussian processes, logistic regression, logit boost, model tree, naive Bayes, nearest neighbour, partial least squares (PLS), random forest, Ridge regression, and support vector machine.
A free 14 day trial of 11Ants Model Builder can be downloaded from www.11antsanalytics.com