A graph based regression model from flat unstructured dataset. Each line in the input data set is treated as a node from which an edge to another line (node) can be formed. In the training process, a model is created which contains sparse graph adjacency matrix. This model is then used for prediction by taking a predictor and the model as inputs and outputs a prediction which is an average of the most similar node and its neighbours in the model graph.
|Date of publication||2016-07-27 11:57:38|
|Maintainer||Yossi Keshet <firstname.lastname@example.org>|