Efficient Disease Classifier Using Data Mining Techniques: Refinement of Random Forest Termination Criteria
Journal Title: IOSR Journals (IOSR Journal of Computer Engineering) - Year 2013, Vol 14, Issue 5
Abstract
In biomedical field, the classification of disease using data mining is the critical task. The prediction accuracy plays a vital role in disease data set. More data mining classification algorithms like decision trees, neural networks, Bayesian classifiers are used to diagnosis the diseases. In decision tree Random Forest, Initially a forest is constructed from ten tress. The accuracy is measured and compared with desired accuracy. If the selected best split of trees matched the desired accuracy the construction terminates. Otherwise a new tree is added with random forest and accuracy is measured. The fitting criteria of random forest are accuracy and correlation. The accuracy is based on the mean absolute percentage error (MAPE) and the mean absolute relative error (MARE).In proposed system to refine the termination criteria of Random Forest, Binomial distribution, multinomial distribution and sequential probability ratio test (SPRT) are used. The proposed method stops the random forest earlier compared with existing Random Forest algorithm. The supervised learning model like support vector machine takes a set of inputs and analyze the inputs and recognize the desired patterns. The disease data sets are supplied to SVM and prediction accuracy is measured. The comparison is made between Random Forest and SVM and best class labels are identified based on disease.
Authors and Affiliations
K. Kalaiselvi
Keywords
Related Articles
- EP ID EP93784
- DOI -
- Views 110
- Downloads 0
Fixed-Rank Representation for Image Classification Using ANN and KNN
Abstract: This study is focused on development an application of an efficient algorithm for subspace clustering and feature extraction that used for classification. Some previous techniques for subspace clustering...
Prototyping and Simulation of Robot Group Intelligence using Kohonen Networks
Abstract:Intelligent agents such as robots can form ad hoc networks and replace human being in many dangerous scenarios such as a complicated disaster relief site. This project prototypes and builds a computer simulator...
An Improved Hashing Method for the Detection of Image Forgery
Abstract: Detection of image forgery is always a crucial factor in image forensic and security applications. Usually this detection is possible with the help of local or global features of an image. We can ensure...
Adaptive Steganography Based Enhanced Cipher HidingTechnique for Secure Data Transfer
Abstract:There have been enormous number of attacks recorded during electronic transmission of informationbetween the source and intended receiver and indeed this has called for a more robust and efficient method f...
Main Content Mining from Multi Data Region Deep Web Pages
Abstract: An Automatic Data extraction from the deep web pages is an endeavoring errand. Colossal web substance are gotten to as per well-known interest submitted to Web databases and the returned data records are e...