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 published on a yearly basis
Papers
More filters
••
05 Sep 2008TL;DR: The authors developed a decision tree classification method that improves the speed and precision of classification and discussed the wild use of LULC decision tree classified and stratified extractive technology.
Abstract: Adopting the decision tree technology, utilizing its process pattern that imitates human judgment and thinking and fault-tolerance features, the authors developed a decision tree classification method. Initially utilizing SPOT and TM, the work effectively enhanced LULC information and established the synthetic database; then, combining geoscience synthetic analysis with ground spectral feature information, utilizing the CART system; the authors built the decision tree model that is based on the decision rules. At last, we discussed the wild use of LULC decision tree classified and stratified extractive technology. Taking three counties in Hebei province as examples, we divided the research area to classify each unit (county area) by ecological division, utilized multiple data resources and geoscience rules to build the decision tree model and test and verify the method. The results demonstrated that the method improves the speed and precision of classification.
3 citations
••
19 Oct 2011TL;DR: The systematic and scientific model for analyzing science technologies and forecasting the emerging technologies in this papers is proposed and can get accuracy of 84%.
Abstract: It becomes important to discover technical opportunity when we due to uncertainties about forecast. As the internet grows rapidly and the amount of information in the web increases exponentially, however, analysis and forecast regarding the science and technology become more difficult. Because decision of emerging technology needs much cost and time, we need more effective method and solution for decision making of prospective science technologies. For overcoming the above limitations, many methods based on non-systemic processes such as Delphi and Scenario technique was suggested. However, the solutions based on non-systemic processes can not sure accuracy of results and show inconsistent forecasting about the science technology. Therefore we propose the systematic and scientific model for analyzing science technologies and forecasting the emerging technologies in this papers. We obtain features using existing technology lifecycle model and make decision tree model consisted of extracted features. In order to evaluate this model, we did performance test toward 50 technologies in Gartner's Hype cycle for emerging technologies 2009~2010, and can get accuracy of 84%.
3 citations
••
10 Oct 2011TL;DR: A novel adjustable reduced metric-first sphere decoding algorithm for tree search that can reduce the complexity of the MFSD and the performance and complexity can be adjusted according QoS.
Abstract: Tree search is an important kind of detection method. It can be used in multiple-input multiple-output systems detection and the multi-user detection. In this paper, a novel adjustable reduced metric-first sphere decoding algorithm for tree search has been proposed. By setting a suitable and adjustable threshold, the proposed algorithm cuts nodes whose partial Euclidean distances are larger than it. Simulation results show that the algorithm can reduce the complexity of the MFSD and the performance and complexity can be adjusted according QoS.
3 citations
•
IBM1
TL;DR: A novel variant of the SH algorithm (MeSH), that uses meta-regressors to determine which candidate configurations should be eliminated at each round, is proposed and applied to the problem of tuning the hyperparameters of a gradient-boosted decision tree model.
Abstract: In this short paper we investigate whether meta-learning techniques can be used to more effectively tune the hyperparameters of machine learning models using successive halving (SH). We propose a novel variant of the SH algorithm (MeSH), that uses meta-regressors to determine which candidate configurations should be eliminated at each round. We apply MeSH to the problem of tuning the hyperparameters of a gradient-boosted decision tree model. By training and tuning our meta-regressors using existing tuning jobs from 95 datasets, we demonstrate that MeSH can often find a superior solution to both SH and random search.
3 citations