Decision tree model

About: Decision tree model is a research topic. Over the lifetime, 2256 publications have been published within this topic receiving 38142 citations.

Specifying and detecting topological changes to an areal object

Abstract: Topological changes to regions, such as merging, splitting, hole formation and elimination, are significant events in the evolution of regions. Wireless sensor network technology, which provides real-time information about the environment, can play an important role in detecting and reporting such topological changes. This thesis provides theoretical foundations and algorithmic solutions to topological change detection using sensor networks. Two models, the morphism-based model and the local tree model, are developed, providing formal semantics of topological changes. The morphism-based model represents dynamic topological properties of continuously evolving areal objects, in which basic and complex topological changes are represented and classified using trees and structure-preserving mappings between them. Based on this model, this work constructs a normal form and proves that it is the simplest form that could represent all the changes under consideration. The local tree model represents discrete and incremental changes of the areal objects based on selected components and relations between them. It allows us to specify different kinds of topological changes using information within the locality of the change. Based on the local tree model, we develop two decentralized and energy-efficient approaches, the transient group-based (TG-based) and the adaptive group-based (AG-based) approaches, to topological change detection using sensor networks. The TG-based approach employs the boundary group framework, which reduces the communication cost by reporting only the group level data instead of data from each individual node. The AG-based approach further reduces the communication cost by reusing the time-invariant information. Experimental results show that when the configurations of sensor networks satisfy certain density and communication constraints, the proposed approaches are able to generate correct reports on the topological changes, and at the same time reduce the communication cost to a level much lower than that of a basic boundary-based data collection approach.

The analysis of human judgment accuracy using decision tree models

Abstract: In this study, a classification model for predicting human judgment accuracy in Taiwan Stock Exchange Capitalization Weighted Stock Index (TAIEX) was developed using the decision tree model. The training data for decision tree learning was obtained from the questionnaire responses of 22 participants in various fields. Experimental results showed the proposed decision tree model is able to explore the important factors that affect forecasters' judgment accuracy. These factors including accumulated gain/loss in investment, education level, occupation type, and working experience, which can be used as criteria for evaluating the accuracy of professional consultants.

Cover Tree Bayesian Reinforcement Learning

Abstract: This paper proposes an online tree-based Bayesian approach for reinforcement learning. For inference, we employ a generalised context tree model. This defines a distribution on multivariate Gaussian piecewise-linear models, which can be updated in closed form. The tree structure itself is constructed using the cover tree method, which remains efficient in high dimensional spaces. We combine the model with Thompson sampling and approximate dynamic programming to obtain effective exploration policies in unknown environments. The flexibility and computational simplicity of the model render it suitable for many reinforcement learning problems in continuous state spaces. We demonstrate this in an experimental comparison with least squares policy iteration.

On the decision complexity of the bounded theories of trees

Abstract: The theory of nite trees is the full rst-order theory of equality in the Herbrand universe (the set of ground terms) over a functional signature containing non-unary function symbols and constants. Albeit decidable, this theory turns out to be of non-elementary complexity [Vor96a]. To overcome the intractability of the theory of nite trees, we introduce in this paper the bounded theory of nite trees. This theory replaces the usual equality =, interpreted as identity, with the in nite family of approximate equalities \down to a xed given depth" f=dgd2!, with d written in binary, and s =d t meaning that the ground terms s and t coincide if all their branches longer than d are cut o . By using a re nement of Ferrante-Racko 's complexitytailored Ehrenfeucht-Fra ss e games, we demonstrate that the bounded theory of nite trees can be decided within linear double exponential space 22cn (n is the length of input) for some constant c > 0.

Inducing NNC-Trees Quickly

Abstract: An NNC-tree is a decision tree (DT) with each non-terminal node containing a nearest neighbor classifier (NNC). Compared with the axis-parallel decision trees (APDTs), NNC-trees are more comprehensible for large problems, because the decision rules corresponding to the trees are simpler. Currently, the author has proposed an algorithm for inducing NNC-trees based on the R4-rule. However, compared with C4.5, which is a popular program for inducing APDTs, the computation of our algorithm is relatively expensive. This paper proposes two methods for reducing the computational cost. The efficiency of the proposed methods is verified through experiments on three public databases.