Data Analysis Computer System and Method Employing Local to Global Causal Discovery
OwnerKonstantinos (Constantin) F. Aliferis, Alexander Statnikovinventors
Life of this patent
Estimating the term…
Abstract
Discovery of causal networks is essential for understanding and manipulating complex systems in numerous data analysis application domains. Several methods have been proposed in the last two decades for solving this problem. The inventive method uses local causal discovery methods for global causal network learning in a divide-and-conquer fashion. The usefulness of the invention is demonstrated in data capturing characteristics of several domains. The inventive method outputs more accurate networks compared to other discovery approaches.
Claims7
A computer-implemented method and system for global causal network discovery from a dataset, comprising the following steps:
- 1) using a configuration of the Generalized Local Learning-Parents and Children (GLL-PC) method, finding the set of parents and children PC(X) for every variable X in the dataset and during this process, the variables may be ordered (for learning their parents and children sets) according to some prioritization sub-process;
- 2) obtaining a graph (skeleton) by representing membership of variables Y in the set PC(X) as undirected edges between Y and X;
- 3) orienting edges in the graph using any method of choice;
- and
- 4) outputting the global causal graph.
A computer-implemented method and system for global causal network discovery from a dataset, comprising the following steps:
- 1) using the Semi-Interleaved HITON-PC method, finding the set of parents and children PC(X) for every variable X in the dataset, and during this process, ordering the variables (for learning their parents and children sets) lexicographically by order of their index in the dataset;
- 2) obtaining a graph (skeleton) as follows:
- adding an edge between X and Y if Y belongs to PC(X) or X belongs to PC(Y);
- 3) orienting edges using greedy steepest-ascent Tabu search and BDeu score;
- and
- 4) outputting the global causal graph.
Dependencies are inferred from the claim text.