Performance Evaluation of Combined Approach of Topic Model and LMS Estimation for Multi-label Classification
Salah Aziz Alamri, Iwao Fujino · Institutional Repositories DataBase (IRDB) · 2018
In this paper, we will give a performance evaluation to a two-stage approach for multi-label classification.At first, we will provide a simple description to our two-stage approach combining topic model and LMS estimation for providing multiple labels to text documents and other kinds of feature data.The first stage of the approach applies unsupervised learning with topic model to obtain a topic distribution for given instances, while the second stage performs supervised learning using the results of the first stage as features.Then we will show some experiment results to evaluate the performance of this combined approach using several typical evaluation open data set for multi-label classification.The results of these experiments confirm that our approach works effectively as expected.