HomeConsultingToolboxesProject WorkAbout UsContact Us

PANDA

Digital data is everywhere -- in every sector, economy, and organization. The ability to store, aggregate, and use it to perform deep analyses can create significant value, and enhance productivity and competitiveness of every organization.
     
Machine learning technologies provide the ability to unlock value and monetize an organization's data assets. However, it is difficult for end users to use machine learning effectively and exploit state-of-the-art methods without support.
     
PANDATM (Predictive Advanced Nonlinear Diagnostic Analyzer) is a set of tools and processes built on scalable technology that solves a wide range of challenging real-world problems, such as classification, ranking, detection, clustering and semi-supervised learning, feature selection, collaborative filtering and adaptation. PANDA utilizes specialized and novel techniques in line with state-of-the-art developments. PANDA is designed for use in a MATLAB environment and offers several outstanding features.
     

  • The capability of any predictive and analytic modeling tools depends on the inclusion of accurate and diverse models in its underlying ensemble of learners. PANDA contains a wide collection of standard learners such as linear and nonlinear support vector machines, regularized logistic regression, bagged or boosted trees, and kernel kNN. More importantly, PANDA includes highly specialized learning models based on advanced optimization technology that can effectively handle nonlinear and interaction variable sections, highly unbalanced or rare class learning, and semi- and unsupervised learning.
      
  • While most off-the-shelf predictive and analytic modeling tools cannot learn complex interactions of input variables, PANDA can effectively handle the nonlinear and interaction variable selection. It uses advanced optimization technology to solve a complex mathematical problem that identifies the most important subset of variables.
      
  • Highly unbalanced datasets can introduce significant bias and present immense computational difficulties to the currently available predictive and analytic modeling tools. PANDA employs highly specialized rare-class learning algorithms that produce high-quality rankings with the ability to scale to very large datasets.
   
  • PANDA can take advantage of unbalanced samples to produce significantly improved models. 
     
We understand that every organization and every problem is different. We work with you through the whole data analysis life-cycle to deliver and deploy a successful solution. If you are interested in using PANDA or working with us, please contact us at cayuga@cayugaresearch.com.