Switched decision tree for simultaneous learning of multiple datasets

This paper proposes a framework to learn multiple datasets simultaneously using a single full decision tree structure. The threshold values are changed with respect to the dataset and stored in a matrix called "mask". Therefore, the full-tree model is called as a switched decision tree. First of all, solvable version of the problem is studied in order to find the decision tree parameters using a genetic algorithm. Then, the proposed algorithm is adapted for a real dataset. Obtained results demonstrate the usefulness of the algorithm for representing multiple datasets within a single switched tree structure.