Classifying uncertain unlabeled data
Publish place: 3rd International Conference on Applied Research in Computer Engineering and Information Technology
Publish Year: 1394
نوع سند: مقاله کنفرانسی
زبان: English
View: 676
This Paper With 5 Page And PDF Format Ready To Download
- Certificate
- من نویسنده این مقاله هستم
استخراج به نرم افزارهای پژوهشی:
شناسه ملی سند علمی:
CITCONF03_246
تاریخ نمایه سازی: 12 تیر 1395
Abstract:
Uncertain data are inherent in many applications, e.g., sensor networks, biometric identification and etc. Also, since the labeling process is difficult and time-consuming for huge data sets, labeling unlabeled data is very important. The aim of this paper is to present an efficient algorithm for classifying unlabeled numerical certain and uncertain data. We extend EM algorithm to label uncertain unlabeled data. The EM algorithm is an iterative approach to maximum likelihood parameter estimation and needs initial parameters. The correct choice of the initial parameters is crucial for convergence of EM algorithm. We use bivariate normal distribution for uncertain data and choose initial parameters based on uncertain labeled data. After estimating the parameters and creating new labeled instances, we classify them using Naive Bayes classification. A comparison of the initial parameters estimation of EM algorithm using common method with the presented method is given. The experimental results show reasonably good agreement with another method.
Keywords:
Classification , Uncertain interval data , uncertain numerical attribute (UNA , Maximum-likelihood estimation (MLE) , Expectation-Maximization (EM) algorithm , Naive Bayes Classification
Authors
Farzad Eskandari.
Department of Statistics, Allameh Tabatabaei University, Tehran, Iran
Imaneh Khodayari Samghabadi.
Ghiaseddin Jamshid Kashani Higher Education Institute
مراجع و منابع این Paper:
لیست زیر مراجع و منابع استفاده شده در این Paper را نمایش می دهد. این مراجع به صورت کاملا ماشینی و بر اساس هوش مصنوعی استخراج شده اند و لذا ممکن است دارای اشکالاتی باشند که به مرور زمان دقت استخراج این محتوا افزایش می یابد. مراجعی که مقالات مربوط به آنها در سیویلیکا نمایه شده و پیدا شده اند، به خود Paper لینک شده اند :