Loading...
Paper Type
Complete
Abstract
The study proposes new neural networks (maps) called SOPM (Self Organizing Prediction Maps) which are a modification of the classic Kohonen Self Organizing Maps (SOM). The SOPM model performs cluster analysis along with its visualization on a two-dimensional map of neurons, however, with the possibility of the distinction of one feature that is represented on one of the map's axes and can be treated as a dependent variable. After completing the learning process, the SOPM model can be used to classify new patterns, as well as in the process of forecasting of the dependent variable together with a visual assessment of the degree of forecast certainty. The benefits of using SOPM were shown on the example of data related to the real estate market (Boston data set).
Recommended Citation
Morajda, Janusz and Paliwoda-Pękosz, Grażyna, "An Enhancement of Kohonen Neural Networks for Predictive Analytics: Self-Organizing Prediction Maps" (2020). AMCIS 2020 Proceedings. 6.
https://aisel.aisnet.org/amcis2020/ai_semantic_for_intelligent_info_systems/ai_semantic_for_intelligent_info_systems/6
An Enhancement of Kohonen Neural Networks for Predictive Analytics: Self-Organizing Prediction Maps
The study proposes new neural networks (maps) called SOPM (Self Organizing Prediction Maps) which are a modification of the classic Kohonen Self Organizing Maps (SOM). The SOPM model performs cluster analysis along with its visualization on a two-dimensional map of neurons, however, with the possibility of the distinction of one feature that is represented on one of the map's axes and can be treated as a dependent variable. After completing the learning process, the SOPM model can be used to classify new patterns, as well as in the process of forecasting of the dependent variable together with a visual assessment of the degree of forecast certainty. The benefits of using SOPM were shown on the example of data related to the real estate market (Boston data set).
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.