A Meta-Top-Down Method for Large-Scale Hierarchical Classification

A META-TOP-DOWN METHOD FOR LARGE-SCALE HIERARCHICAL CLASSIFICATION

ABSTRACT:

Recent large-scale hierarchical classification tasks typically have tens of thousands of classes on which the most widely used approach to multiclass classification—one-versus-rest—becomes intractable due to computational complexity. The top-down methods are usually adopted instead, but they are less accurate because of the so-called error-propagation problem in their classifying phase. To address this problem, this paper proposes a meta-top-down method that employs meta classification to enhance the normal top-down classifying procedure. The proposed method is first analyzed theoretically on complexity and accuracy, and then applied to five real-world large-scale data sets. The experimental results indicate that the classification accuracy is largely improved, while the increased time costs are smaller than most of the existing approaches.

EXISTING SYSTEM:

The ensemble method of one-versus-rest is the most widely adopted solution for multiclass classification. First a binary-class classifier fi ði ¼ 1; . . . ; nÞ, named base classifier, is trained for each class ci to predict whether an input sample x belongs to this class; then thresholding strategies are employed to decide the predicted labels according to the confidence scores of the base classifiers. Two commonly used thresholding strategies are score-cut (S-cut) that accepts the classes whose scores are larger than a predefined threshold, and rank-cut (R-cut) that accepts the classes whose scores are among the top-r (r is a predefined integer). In recent years, two types of classification strategies have been developed for multiclass classification problems.

DISADVANTAGES OF EXISTING SYSTEM:

v Large numbers of predefined classes.

v It becomes intractable due to computational complexity.

v Error propagation.

v Rejecting a child node at high layers a blocking problem.

PROPOSED SYSTEM:

The proposed MetaTD employs Meta classification to fuse the scores of all the base classifiers instead of making arbitrary cascaded decisions. The framework is to encode the scores along

a root-to-leaf path into a feature vector, and to employ a meta classifier to predict whether the corresponding leaf node is a correct label. Note that there are two kinds of hierarchical classification tasks in real-world applications. One kind is mandatory leaf-node classification, where only the leaf nodes are valid labels. In contrast, the other is non mandatory leaf-node classification, where both the internal nodes and the leaf nodes are valid labels.

ADVANTAGES OF PROPOSED SYSTEM:

v Having meta-top-down method for large-scale hierarchical classification.

v It provides Provide accuracy and complexity analysis.

SYSTEM CONFIGURATION:-

HARDWARE REQUIREMENTS:-

Processor                  -        Pentium –IV

Speed                        -        1.1 Ghz

RAM                         -        512 MB(min)

Hard Disk                 -        40 GB

Key Board                -        Standard Windows Keyboard

Mouse                       -        Two or Three Button Mouse

Monitor                     -        LCD/LED

SOFTWARE REQUIREMENTS:

Operating system      :         Windows XP.

Coding Language      :         .Net

Data Base                 :         SQL Server 2005

Tool                          :         VISUAL STUDIO 2008.

REFERENCE:

A Meta-Top-Down Method for Large-Scale Hierarchical Classification, “A Meta-Top-Down Method for Large-Scale Hierarchical Classification” IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING, VOL. 26, NO. 3, MARCH 2014