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
11 Mar 2014
TL;DR: The experimental results showed that multi-role hidden tree model can also produce scale-free networks, indicating that thehidden tree model is robust with multiple roles.
Abstract: Based on the spatial factor other than the temporal accumulation, the hidden tree model was built up to model scale-free networks. This paper further assumed that a node has multiple roles in different hidden trees, and explored the multi-role hidden tree model. The experimental results showed that multi-role hidden tree model can also produce scale-free networks. This conclusion indicates that the hidden tree model is robust with multiple roles.
08 May 2017
TL;DR: A probabilistic graphical model of ranked preferences for social choice based on a restricted version of a kth-order Markov tree based on the intuition that, in some domains, an agent's next most preferred alternative is highly predictable given a few of the immediately preceding alternatives is introduced.
Abstract: We introduce a probabilistic graphical model of ranked preferences for social choice based on a restricted version of a kth-order Markov tree. The system is similar to Plackett's model, and is based on the intuition that, in some domains, an agent's next most preferred alternative is highly predictable given a few of the immediately preceding alternatives. We provide a bound on the data requirements to learn the parameters of such a model and show eventual consistency between most probable sequences of the model and both the centroid of a Mallows model and the induced ranking of a RUM, theoretically and on artificial data; this is followed by a full evaluation of the model as a social choice system, including comparisons with existing approaches on 6 real world datasets. An application of the system to a simulated multiagent coordination task is also demonstrated, in which the proposed system offers pronounced advantages over existing approaches.
Patent•
26 Jun 2020
TL;DR: In this paper, a method for training endometriosis cyst rupture data based on a random forest algorithm was proposed, and the method comprises the steps: obtaining endometriaosis cysts rupture and non-rupture data as sample data, carrying out the normalization processing, and dividing the sample data into a test set and a plurality of training sets; performing decision tree training on each training set; selecting an optimal feature from each CART decision tree model through Gini index comparison to obtain a corresponding decision tree and form a random tree model; performing parameter optimization on
Abstract: The invention provides a method for training endometriosis cyst rupture data based on a random forest algorithm, and the method comprises the steps: obtaining endometriosis cyst rupture and non-rupture data as sample data, carrying out the normalization processing, and dividing the sample data into a test set and a plurality of training sets; performing decision tree training on each training setto obtain a corresponding CART decision tree model; selecting an optimal feature from each CART decision tree model through Gini index comparison to perform branching processing to obtain a corresponding decision tree and form a random forest model; performing parameter optimization on the random forest model by adopting a particle swarm algorithm, and importing the random forest model into the training set and the test set to obtain a trained random forest model; and acquiring endometriosis cyst data to be detected, importing the data into the trained random forest model, and distinguishing rupture or non-rupture data. By implementing the method, a continuous, discrete and mixed endometriosis cyst rupture data set can be processed, and the problem that the accuracy is quickly reduced under the condition of lack of more data is solved.
TL;DR: In this paper, a method called Clinical Heart Disease-Decision Supportive Optimized Mining (CHDDSOM) is presented to overcome the pattern matching loss, and perform perfect pattern matching.
Abstract: Healthcare industry collects enormous amounts of healthcare data and required to mine and ascertain hidden information for constructive decision making. In recent years, the computer technology and machine learning methods with data mining approach increase techniques in assisting the doctors for productive decision making related to heart disease and stroke identification at an early stage. The needs to reduce the pattern matching loss by performing pattern matching while effectively improving the heart disease identification at much early stage poses severe challenges to the database community. In this paper a method called Clinical Heart Disease-Decision Supportive Optimized Mining (CHDDSOM) is presented to overcome the pattern matching loss, and perform perfect pattern matching. The method CHDDSOM categories to enable decision support with multidimensional analysis using Mahalanobis distance measure for obtaining dynamic data table information and typically built to support early stage of stroke identification and levels of heart disease. This helps in identifying the solution for different patterns and therefore reducing the pattern matching loss. With the objective of identifying the level of heart disease in a very accurate manner, the CHDDSOM method uses the Iterative Dichotomiser 3 based decision tree model.With Iterative Dichotomiser 3 based decision tree model, the stroke is also identified very easily by starting with the original set S as the root node. The entropy value of every attribute is calculated and is placed in the decision tree. Decision tree are constructed from the higher to lower values to perform easy pattern matching, aiming at reducing the processing time for pattern matching. Experiment is conducted with the Cleveland Clinic Foundation Heart disease data set available from UCI repository using the factors such as pattern matching loss rate, accuracy, processing time for pattern matching, computational cost. Experimental analysis shows that the CHD-DSOM method is able to reduce the processing time for pattern matching by 33.23% and reduce the pattern matching loss rate by 29.67% compared to the state-of-the-art works.
TL;DR: In this article , decision tree models are developed on dispersed data using entropy measure and twoing criterion as the splitting criteria, and the main purpose of this paper is to make a comparative study on the classification quality of decision tree model built on dispersed data using twoing splitting measure.