A Practical Approach to Classify Evolving Data Streams: Training with Limited Amount of Labeled Data. Masud, M. M., Gao, J., Khan, L., Han, J., & Thuraisingham, B. In 2008 Eighth IEEE International Conference on Data Mining, pages 929–934, December, 2008. ISSN: 2374-8486doi abstract bibtex Recent approaches in classifying evolving data streams are based on supervised learning algorithms, which can be trained with labeled data only. Manual labeling of data is both costly and time consuming. Therefore, in a real streaming environment, where huge volumes of data appear at a high speed, labeled data may be very scarce. Thus, only a limited amount of training data may be available for building the classification models, leading to poorly trained classifiers. We apply a novel technique to overcome this problem by building a classification model from a training set having both unlabeled and a small amount of labeled instances. This model is built as micro-clusters using semi-supervised clustering technique and classification is performed with kappa-nearest neighbor algorithm. An ensemble of these models is used to classify the unlabeled data. Empirical evaluation on both synthetic data and real botnet traffic reveals that our approach, using only a small amount of labeled data for training, outperforms state-of-the-art stream classification algorithms that use twenty times more labeled data than our approach.
@inproceedings{masud_practical_2008,
title = {A {Practical} {Approach} to {Classify} {Evolving} {Data} {Streams}: {Training} with {Limited} {Amount} of {Labeled} {Data}},
shorttitle = {A {Practical} {Approach} to {Classify} {Evolving} {Data} {Streams}},
doi = {10.1109/ICDM.2008.152},
abstract = {Recent approaches in classifying evolving data streams are based on supervised learning algorithms, which can be trained with labeled data only. Manual labeling of data is both costly and time consuming. Therefore, in a real streaming environment, where huge volumes of data appear at a high speed, labeled data may be very scarce. Thus, only a limited amount of training data may be available for building the classification models, leading to poorly trained classifiers. We apply a novel technique to overcome this problem by building a classification model from a training set having both unlabeled and a small amount of labeled instances. This model is built as micro-clusters using semi-supervised clustering technique and classification is performed with kappa-nearest neighbor algorithm. An ensemble of these models is used to classify the unlabeled data. Empirical evaluation on both synthetic data and real botnet traffic reveals that our approach, using only a small amount of labeled data for training, outperforms state-of-the-art stream classification algorithms that use twenty times more labeled data than our approach.},
booktitle = {2008 {Eighth} {IEEE} {International} {Conference} on {Data} {Mining}},
author = {Masud, Mohammad M. and Gao, Jing and Khan, Latifur and Han, Jiawei and Thuraisingham, Bhavani},
month = dec,
year = {2008},
note = {ISSN: 2374-8486},
keywords = {Buffer storage, Classification algorithms, Clustering algorithms, Computer science, Data mining, Data stream, Labeling, Probability distribution, Supervised learning, Testing, Training data, ensemble classification, semi-supervised clustering},
pages = {929--934},
}
Downloads: 0
{"_id":"JcCaTBSYPdFBTM9ih","bibbaseid":"masud-gao-khan-han-thuraisingham-apracticalapproachtoclassifyevolvingdatastreamstrainingwithlimitedamountoflabeleddata-2008","author_short":["Masud, M. M.","Gao, J.","Khan, L.","Han, J.","Thuraisingham, B."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","title":"A Practical Approach to Classify Evolving Data Streams: Training with Limited Amount of Labeled Data","shorttitle":"A Practical Approach to Classify Evolving Data Streams","doi":"10.1109/ICDM.2008.152","abstract":"Recent approaches in classifying evolving data streams are based on supervised learning algorithms, which can be trained with labeled data only. Manual labeling of data is both costly and time consuming. Therefore, in a real streaming environment, where huge volumes of data appear at a high speed, labeled data may be very scarce. Thus, only a limited amount of training data may be available for building the classification models, leading to poorly trained classifiers. We apply a novel technique to overcome this problem by building a classification model from a training set having both unlabeled and a small amount of labeled instances. This model is built as micro-clusters using semi-supervised clustering technique and classification is performed with kappa-nearest neighbor algorithm. An ensemble of these models is used to classify the unlabeled data. Empirical evaluation on both synthetic data and real botnet traffic reveals that our approach, using only a small amount of labeled data for training, outperforms state-of-the-art stream classification algorithms that use twenty times more labeled data than our approach.","booktitle":"2008 Eighth IEEE International Conference on Data Mining","author":[{"propositions":[],"lastnames":["Masud"],"firstnames":["Mohammad","M."],"suffixes":[]},{"propositions":[],"lastnames":["Gao"],"firstnames":["Jing"],"suffixes":[]},{"propositions":[],"lastnames":["Khan"],"firstnames":["Latifur"],"suffixes":[]},{"propositions":[],"lastnames":["Han"],"firstnames":["Jiawei"],"suffixes":[]},{"propositions":[],"lastnames":["Thuraisingham"],"firstnames":["Bhavani"],"suffixes":[]}],"month":"December","year":"2008","note":"ISSN: 2374-8486","keywords":"Buffer storage, Classification algorithms, Clustering algorithms, Computer science, Data mining, Data stream, Labeling, Probability distribution, Supervised learning, Testing, Training data, ensemble classification, semi-supervised clustering","pages":"929–934","bibtex":"@inproceedings{masud_practical_2008,\n\ttitle = {A {Practical} {Approach} to {Classify} {Evolving} {Data} {Streams}: {Training} with {Limited} {Amount} of {Labeled} {Data}},\n\tshorttitle = {A {Practical} {Approach} to {Classify} {Evolving} {Data} {Streams}},\n\tdoi = {10.1109/ICDM.2008.152},\n\tabstract = {Recent approaches in classifying evolving data streams are based on supervised learning algorithms, which can be trained with labeled data only. Manual labeling of data is both costly and time consuming. Therefore, in a real streaming environment, where huge volumes of data appear at a high speed, labeled data may be very scarce. Thus, only a limited amount of training data may be available for building the classification models, leading to poorly trained classifiers. We apply a novel technique to overcome this problem by building a classification model from a training set having both unlabeled and a small amount of labeled instances. This model is built as micro-clusters using semi-supervised clustering technique and classification is performed with kappa-nearest neighbor algorithm. An ensemble of these models is used to classify the unlabeled data. Empirical evaluation on both synthetic data and real botnet traffic reveals that our approach, using only a small amount of labeled data for training, outperforms state-of-the-art stream classification algorithms that use twenty times more labeled data than our approach.},\n\tbooktitle = {2008 {Eighth} {IEEE} {International} {Conference} on {Data} {Mining}},\n\tauthor = {Masud, Mohammad M. and Gao, Jing and Khan, Latifur and Han, Jiawei and Thuraisingham, Bhavani},\n\tmonth = dec,\n\tyear = {2008},\n\tnote = {ISSN: 2374-8486},\n\tkeywords = {Buffer storage, Classification algorithms, Clustering algorithms, Computer science, Data mining, Data stream, Labeling, Probability distribution, Supervised learning, Testing, Training data, ensemble classification, semi-supervised clustering},\n\tpages = {929--934},\n}\n\n","author_short":["Masud, M. M.","Gao, J.","Khan, L.","Han, J.","Thuraisingham, B."],"key":"masud_practical_2008","id":"masud_practical_2008","bibbaseid":"masud-gao-khan-han-thuraisingham-apracticalapproachtoclassifyevolvingdatastreamstrainingwithlimitedamountoflabeleddata-2008","role":"author","urls":{},"keyword":["Buffer storage","Classification algorithms","Clustering algorithms","Computer science","Data mining","Data stream","Labeling","Probability distribution","Supervised learning","Testing","Training data","ensemble classification","semi-supervised clustering"],"metadata":{"authorlinks":{}}},"bibtype":"inproceedings","biburl":"https://bibbase.org/zotero/mh_lenguyen","dataSources":["iwKepCrWBps7ojhDx"],"keywords":["buffer storage","classification algorithms","clustering algorithms","computer science","data mining","data stream","labeling","probability distribution","supervised learning","testing","training data","ensemble classification","semi-supervised clustering"],"search_terms":["practical","approach","classify","evolving","data","streams","training","limited","amount","labeled","data","masud","gao","khan","han","thuraisingham"],"title":"A Practical Approach to Classify Evolving Data Streams: Training with Limited Amount of Labeled Data","year":2008}