scispace - formally typeset
Search or ask a question
Topic

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.


Papers
More filters
Patent
13 Apr 2016
TL;DR: Wang et al. as discussed by the authors proposed a tree structure based sorting method for processing Chinese webpage data in a Chinese search engine, which comprises: step S100, preprocessing webpage data, step S200, establishing a webpage data index file; step S300, receiving a query character string input by a user, and performing retrieval according to the webpage data; and step S400, sorting retrieval results and displaying a sorting result to the user.
Abstract: The invention proposes a tree structure based sorting method. The method is used for processing Chinese webpage data in a Chinese search engine. The method comprises: step S100, preprocessing webpage data; step S200, establishing a webpage data index file; step S300, receiving a query character string input by a user, and performing retrieval according to the webpage data; and step S400, sorting retrieval results and displaying a sorting result to the user. According to the method, indexes are created for the webpage data by adopting a binary inter-correlative subsequent tree model, and the advantages and disadvantages of word indexing and phrase indexing are considered, so that the retrieval efficiency is improved while the index space is reduced.
Patent
12 Jan 2018
TL;DR: In this article, an intelligent matching method of a wireless virtualization accessing automatic management network service and a virtual service is proposed, which includes offline study and online recognition, wherein offline study includes: adopting a proximal gradient descent method to perform dimension reduction on features, selecting features of decision tree and establishing a decision feature set A'; performing scope clustering on features xi in the decision feature sets A' by adopting a lagrangian operator; determining priority of the decision features; establishing a tree model according to the priority of decision features, online recognition includes: starting form root nodes
Abstract: The invention relates to an intelligent matching method of a wireless virtualization accessing automatic management network service and a virtual service. The method includes offline study and onlinerecognition, wherein offline study includes: adopting a proximal gradient descent method to perform dimension reduction on features, selecting features of decision tree and establishing a decision feature set A'; performing scope clustering on features xi in the decision feature set A' by adopting a lagrangian operator; determining priority of the decision features; establishing a decision tree model according to the priority of the decision features; online recognition includes: starting form root nodes of the tree, and comparing nodal values of the decision tree model acquired from offline study with values of corresponding domains in the wireless access network packet data protocol; if the values of the domains belong to subdomains of the nodes of the decision tree model acquired from offline study, subjecting the subdomains of the nodes of the decision tree model acquired from offline study to recursion processing till the corresponding virtual services are found. As is proved by practice, the intelligent matching method is high in matching speed, low in misjudgment rate and capable of adapting to migration of different scenarios.
Journal ArticleDOI
TL;DR: A dynamic programming algorithm for converting decision tables to optimal decision trees is analyzed and methods of reducing the dimensionality of the problem utilizing lower bounds for decision costs are discussed.
Abstract: A dynamic programming algorithm for converting decision tables to optimal decision trees is analyzed. The complexity of the algorithm may be defined as the dimension of the domain of the minimal-cost functional. Upper bounds for this complexity are derived under various assumptions. Methods of reducing the dimensionality of the problem utilizing lower bounds for decision costs are also discussed.
Proceedings ArticleDOI
21 Aug 2019
TL;DR: Among all models, decision tree model with Entropy algorithm perform the best by scoring the highest accuracy rate and sensitivity rate and has been selected as the best model for predicting the ability of the Ph.D students in achieving Graduate on Time (GOT).
Abstract: Over the years, there has been exponential growth in the number of Doctor of Philosophy (Ph.D) graduates in most of the universities all around the world. The increment of Ph.D students causes both university and government bodies concern about the capability of the Ph.D students to accomplish the mission of Graduate on Time (GOT) that is stipulated by the university. Therefore, this study aims to classify the Ph.D students into the group of “GOT achiever” and “non-GOT achiever” by using decision tree models. Historical data that related to all Ph.D students in a public university in Malaysia has been obtained directly from the database of Graduate Academic Information System (GAIS) in order to develop and compare the performance of decision tree models (Chi-square algorithm, Gini index algorithm, Entropy algorithm and an interactive decision tree). The result gained in four decision tree models illustrated that the attributes of English background, gender and the Ph.D students’ entry Cumulative Grade Point Average (CGPA) result are the core in impacting the students’ success. Among all models, decision tree model with Entropy algorithm perform the best by scoring the highest accuracy rate (72%) and sensitivity rate (95%). Therefore, it has been selected as the best model for predicting the ability of the Ph.D students in achieving GOT. The outcome can certainly ease the burden of universities in handling and controlling the GOT issue. Also, the model can be used by the university to uncover the restriction in this issue so that better plans can be carried out to boost the number of GOT achiever in future.
Posted Content
03 Sep 2009
TL;DR: A proper tree search technique that reduces overall SD computational complexit y without sacrificing performance, and builds a check-table to pre-calculate and store some terms, and temporally store mid-stage terms, to take advantage of a new lattice representation of previous work.
Abstract: In Multiple-Input Multiple-Output (MIMO) systems, Sphere Decoding (SD) can achieve performance equivalent to full search Maximum Likelihood (ML) decoding, with reduced complexity. Several researchers reported techniques that reduce the complexity of SD further. In this paper, a new technique is introduced which decreases the computational complexity of SD substantially, without sacrificing performance. The reduction is accomplished by deconstructing the decoding metric to decrease the number of computations and exploiting the structure of a lattice representation. Furthermore, an application of SD, employing a proposed smart implementation with very low computational complexity is introduced. This application calculates the soft bit metrics of a bit-interleaved convolutional-coded MIMO system in an efficient manner. Based on the reduced complexity SD, the proposed smart implementation employs the initial radius acquired by Zero-Forcing Decision Feedback Equalization (ZF-DFE) which ensures no empty spheres. Other than that, a technique of a particular data structure is also incorporated to efficiently reduce the number of executions carried out by SD. Simulation results show that these approaches achieve substantial gains in terms of the computational complexity for both uncoded and coded MIMO systems.

Network Information
Related Topics (5)
Cluster analysis
146.5K papers, 2.9M citations
80% related
Artificial neural network
207K papers, 4.5M citations
78% related
Fuzzy logic
151.2K papers, 2.3M citations
77% related
The Internet
213.2K papers, 3.8M citations
77% related
Deep learning
79.8K papers, 2.1M citations
77% related
Performance
Metrics
No. of papers in the topic in previous years
YearPapers
202310
202224
2021101
2020163
2019158
2018121