Mining Decision Tree Research Articles

The Decision Tree and Support Vector Machine for the Data Mining

The Decision Tree and Support Vector Machine for the Data Mining

A novel data mining mechanism considering bio-signal and environmental data with applications on asthma monitoring

Chronic asthmatic sufferers need to be constantly observed to prevent sudden attacks. In order to improve the efficiency and effectiveness of patient monitoring, we proposed in this paper a novel data mining mechanism for predicting attacks of chronic diseases by considering of both bio-signals of patients and environmental factors. We proposed two data mining methods, namely Pattern Based Decision Tree ( PBDT) and Pattern Based Class-Association Rule ( PBCAR). Both methods integrate the concepts of sequential pattern mining to extract features of asthma attacks, and then build classifiers with the concepts of decision tree mining and rule-based method respectively. Besides the general clinical data of patients, we considered environmental factors, which are related to many chronic diseases. For experimental evaluations, we adopted the children asthma allergic dataset collated from a hospital in Taiwan as well as the environmental factors like weather and air pollutant data. The experimental results show that PBCAR delivers 86.89% of accuracy and 84.12% of recall, and PBDT shows 87.52% accuracy and 85.59 of recall. These results also indicate that our methods can perform high accuracy and recall on predictions of chronic disease attacks. The readable rules of both classifiers can provide patients and healthcare workers with insights on essential illness related information. At the same time, additional environmental factors of input data are also proven to be valuable in predicting attacks.

Optimal constraint-based decision tree induction from itemset lattices

In this article we show that there is a strong connection between decision tree learning and local pattern mining. This connection allows us to solve the computationally hard problem of finding optimal decision trees in a wide range of applications by post-processing a set of patterns: we use local patterns to construct a global model. We exploit the connection between constraints in pattern mining and constraints in decision tree induction to develop a framework for categorizing decision tree mining constraints. This framework allows us to determine which model constraints can be pushed deeply into the pattern mining process, and allows us to improve the state-of-the-art of optimal decision tree induction.

Media Consumption in Taiwan: Mining the Shih Hsin Communication Database

Internet use has recently grown rapidly, and this trend is continuing. This increase is unsurprising given the reallocation of attention from newspapers, magazines, radio and television. Since most audiences have limited time available for media use, it is logical to assume that the hours spent using a particular medium reduce or displace the time spent using other media. However, increase-increase in some media reinforcement effects among media may occur according to the bio-ecological theory of the niche. The purpose of this investigation was to better understand media usage in Taiwan. The two most commonly used data mining techniques, neural networks and decision trees, along with some statistical methods, are applied to profile heavy-users of each of the five major media: TV, radio, newspaper, magazine and Internet. The correlations among different media consumption behaviors are summarized using a ”Media Usage Correlation Matrix” in this study. The analytical results of the investigation reveal some interesting findings relating to the five media usage and the relationship among media use. For example, the consumption behavior for printed media (newspapers and magazines) has a complementary effect while the use of newspapers has a displacement effect. Moreover, magazines exert a complementary effect on radio, TV and Internet use. Furthermore, the study demonstrates that heavy users of the five media by constructing data mining models, respectively. The findings of this study offer a potential new direction that media buyers can use to select efficient advertising media vehicles after targeting a specific population.

Predicting consumer preference for fast-food franchises: a data mining approach

The objectives of the study reported in this paper are: (1) to evaluate the adequacy of two data mining techniques, decision tree and neural network in analysing consumer preference for a fast-food franchise and (2) to examine the sufficiency of the criteria selected in understanding this preference. We build decision tree and neural network models to fit data samples collected from 800 respondents in Taiwan to understand the factors that determine their brand preference. Classification rules are generated from these models to differentiate between consumers who prefer the brand and those who do not. The generated rules show that while both decision tree and neural network models can achieve predictive accuracy of more than 80% on the training data samples and more that 70% on the cross-validation data samples, the neural network models compare very favourably to a decision tree model in rule complexity and the numbers of relevant input attributes.

Applying data mining to explore the risk factors of parenting stress

High parenting stress has been shown to be associated with illness, poor marital relationships, and child abuse. It is thus important to detect and reduce the stress early. Numerous situational, socioeconomic, child, and parent factors contribute to parenting stress. However, regression analysis, the traditional method of exploring risk factors in medicine and social science, has the limitation of not showing the classification, nor exploring unknown potentional factors. Hence, the goal of this study is to explore the risk factors of parenting stress using data mining with decision tree C5.0, to obtain more information. The data are from a professional research group, TBPS, in the National Taiwan University. A total of 206 mother-term born child dyads were recruited to complete the measures of the Parenting Stress Index (PSI), the Child Behaviour Checklist (CBCL), the Comprehensive Developmental Inventory for Infants and Toddlers (CDIIT), and the Chinese Toddler Temperament Scale (CTTS), and so this database includes thousands of variables. The study results indicate that a child development problem, CDIIT, is the major contributing factor to parents with the highest stress, the 90% parenting stress group. For the 80%, 70%, and 60% parenting stress groups, the behavioural problem of children, CBCL, is the major factor causing parenting stress. The data mining decision tree showing the classification route of risk factors is better than the regression model at detecting the significant factors. The findings in this work are considered helpful references for medical staff and social workers to help parents prevent and reduce their parenting stress and thus promote health.

Measurement and Management for Medical and Clinical Risks

Hosptial information system (HIS) collects all the data from all the branches of departments in a hospital, including laboratory tests, physiological tests, electronic patient records. Thus, HIS can be viewed as a large heterogenous database, which stores chronological changes in patients' status. This paper overviews three applications of data mining and statistical methods to HIS. First, clustering of temporal sequences based on multiscale matching was applied for grouping chronic hepatitis. Second, decision tree method was used for detection of risk factors, which was successfully used to prevent nursing medication errors. Finally, several linear models were applied for hospital management data. These results show that data mining methods, including decision tree mining, temporal data mining, are useful for detection of risk factors from large distributed data such as HIS, whose process can be called risk mining.

Predicting Breast Cancer Survivability: Comparison of Five Data Mining Techniques

Objective: Today in United States, about one in eight women have been affected with breast cancer over their lifetime. Up to today, some various prediction models using SEER (Surveillance Epidemiology and End Results) datasets have been proposed in past studies. However, appropriate methods for predicting the 5 years survival rate of breast cancer have not established. In this study, we evaluate those models to predict the survival rate of breast cancer patients. Methods: Five data mining algorithms (Artificial Neural Network, Naive Bayes , Decision Trees (ID3) and Decision Trees(J48)) besides a most generally used statistical method (Logistic Regression) were used to evaluate the prediction models using a dataset (37,256 follow-up cases from 1992 to 1997). We also used 10-fold cross-validation methods to assess the unbiased estimate of the five prediction models for comparison of performance of each method. Results: The accuracy was 85.8±0.2%, 84.3±1.4%, 83.9±0.2%, 82.3±0.2%, 75.1±0.2% for the Logistic Regression, Artificial Neural, Naive Bayes, Decision Trees (ID3), Decision Trees(J48), respectively. Although the accuracy of Logistic Regression showed the highest performances, the Decision Trees (J48) was the lowest one. Conclusions: The accuracy of Logistic Regression was the best performances, on the other hand Decision Trees (J48) was the worst. Artificial Neural Network indicated relatively high performance. (Journal of Korean Society of Medical Informatics 13-2, 177-180, 2007)

A Simple Scheme to Determine Potential Aquatic Metal Toxicity from Mining Wastes

A decision tree (mining waste decision tree) that uses simple physical and chemical tests has been developed to determine whether effluent from mine waste material poses a potential toxicity threat to the aquatic environment. For the chemical portion of the tree, leaching tests developed by the United States Geological Survey, the Colorado Division of Minerals and Geology (Denver, CO), and a modified 1311 toxicity characteristic leaching procedure (TCLP) test of the United States Environmental Protection Agency have been extensively used as a surrogate for readily available metals that can be released into the environment from mining wastes. To assist in the assessment, element concentration pattern graphs (ECPG) are produced that compare concentrations of selected groups of elements from the three leachates and any water associated with the mining waste. The MWDT makes a distinction between leachates or waters with pH less than or greater than 5. Generally, when the pH values are below 5, the ECPG of the solutions are quite similar, and potential aquatic toxicity from cationic metals, such as Pb, Cu, Zn, Cd, and Al, is assumed. Below pH 5, these metals are mostly dissolved, generally are not complexed with organic or inorganic ligands, and hence are more bioavailable. Furthermore, there is virtually no carbonate alkalinity at pH less than 5. All of these factors promote metal toxicity to aquatic organisms. On the other hand, when the pH value of the water or the leachates is above 5, the ECPG from the solutions are variable, and inferred aquatic toxicity depends on factors in addition to the metals released from the leaching tests. Hence, leachates and waters with pH above 5 warrant further examination of their chemical composition. Physical ranking criteria provide additional information, particularly in areas where waste piles exhibit similar chemical rankings. Rankings from physical and chemical criteria generally are not correlated. Examples of how this decision tree has been applied in assessing mine sites are discussed.

DECISION TREES FOR PERNICIOUS PAGES DETECTION

An application framework to perform web usage analysis using advanced data mining methodology is presented. The investigation proposes decision trees for web user behavior analysis. This includes prediction of user future actions and the typical pages leading to browsing termination. The widely known decision tree package C4.5 was used in this study. In the new area of web log mining decision trees showed reasonable computational performance and accuracy. Experiments showed that it is possible to predict future user actions with reasonable misclassification error as well as to find combinations of sequential pages resulting in browsing termination. In addition to this, decision trees generate human understandable rules which can be used to analyze further for web site improvement.