Unsupervised Segmentation Methodology

Challenges

 

The analysis of large volumes of data presents the challenge of discovering similar patterns of behaviour, trends or groups who do not respond to an idea predefined by an expert. This requires using unsupervised analysis tools, which are able to structure the information for subsequent analysis.


Approach

 

ADAN is a tool developed with Artificial Intelligence techniques. Unsupervised, it automatically extracts the underlying structure of a database. The groups or clusters are formed by the similarity of the elements or components of the database regarding variables, which are subject to analysis. The system has been designed for the analysis of bulk data. Through innovative techniques, ADAN allows experts to analyse the information available to support the decision making process.

ADAN Clustering is a conceptual tool based on unsupervised learning techniques. It uses a powerful exclusive polythetic algorithm, unsupervised, and fully developed by Grupo AIA. The system allows segmentation completion without the constraints of statistical techniques. It includes topological recognition, a contextual explanatory component, and graphical projection of the resulting spatial clusters, among other notable features.

ADAN is an analysis tool that can be considered for general purposes. Processes can decisively help the definition of a company, establishing growth strategies, financing investment in a network of branches, definition of institutional marketing, and sales strategies. It has been used to classify, along with others, data in order to determine customer profiles for marketing and sales departments, risk analysis for customers in banking scope, Internet surfers and customers of virtual stores, branches for a corporation and product performance for quality control.

 


Benefits

 

  • It addresses the problem of grouping the elements of a database into classes or clusters based on their similarities.
  • Uses unsupervised learning techniques, being a product of artificial intelligence.
  • Polythetic clustering algorithm (a set of variables chosen, taken simultaneously).
  • The advanced internal algorithm allows unsupervised and automatic selection of the number of end groups.
  • Graphical representation of the results as a hierarchical tree.
  • Explanation of branching at any level of the tree.
  • Option groups description for conceptualisation (label placement).
  • Automatic selection of a statistical sample to expedite the subsequent process of clustering.
  • Optional Modification of basic Euclidean metric using supervised learning techniques.
  • Further refinement of the results of a group by defining a group or groups merge as a new universe.
  • Automatic scaling range of each variable, detecting relevant discrete values.
  • Quality control elements of the process: quality index and spatial representation of the data.
  • Option list of components of a group.
  • Identification of outliers and profiling.
  • Option list of dissidents in a group.
  • Tree pruning option.
  • Move Options and modifying the tree size.

Intellectual Property

 

ADAN (Trademark)

Ownership: AIA

Appl. No.: 576.478 & M1.613.360 -9