Self-Organizing Maps (SOMs)The SOM is a type of neural network that was originally developed for data visualization, where the objective is to reduce data from a high number of dimensions (attributes) down to two so they can be seen on a map. The process of reducing the dimensionality of the data also results in grouping based on similarities of the observations. Essentially this means that SOMs segment data when they reduce the dimensionality of the data. Through tests in various industries we have found that segments created by this process are very attractive and actionable and often superior to segments formed by more conventional methods like cluster analysis. SOMs fall under a category called unsupervised neural networks.
|
Register for TRC's latest research seminar - Tuesday, Oct 5th in the Yale Club of New York. Research 2.0: Innovative choice and Social Networks Methods. This one-day event will feature speakers from Columbia and Rice Universities, in addition to our very own Rajan Sambandam and Pankaj Kumar. The event will highlight new and interesting developments in choice techniques, text mining, and "Research 2.0". To learn more and register, click here. |
| TRC's White Paper Library offers, among others, various choice-related articles. Featured Choice-Related White Papers: How to improve your segmentation with Max-Diff Conjoint Analysis vs. Self-explicated Understanding choice in banking: Use of Discrete choice analysis Sign Up to receive TRC e-Newsletters |
