東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁

切換: 標籤 | MARC模式 | ISBD

Mining User Behaviors in Social Netw...

Sayyadiharikandeh, Mohsen.

FindBook

Google Book

Amazon

博客來

Mining User Behaviors in Social Networks Using Machine Learning.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Mining User Behaviors in Social Networks Using Machine Learning./
作者:	Sayyadiharikandeh, Mohsen.
出版者:	Ann Arbor : ProQuest Dissertations & Theses, : 2021,
面頁冊數:	138 p.
附註:	Source: Dissertations Abstracts International, Volume: 82-09, Section: B.
Contained By:	Dissertations Abstracts International82-09B.
標題:	Computer science. -
電子資源:	https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=28317657
ISBN:	9798582516637

Mining User Behaviors in Social Networks Using Machine Learning.
Sayyadiharikandeh, Mohsen.

Mining User Behaviors in Social Networks Using Machine Learning. - Ann Arbor : ProQuest Dissertations & Theses, 2021 - 138 p.

Source: Dissertations Abstracts International, Volume: 82-09, Section: B.

Thesis (Ph.D.)--Indiana University, 2021.

This item must not be sold to any third party vendors.

The work described in my thesis is in the area of applied machine learning. I developed supervisedmethods that leverage user behavior in online social and knowledge networks to infer some specific attributes of such users. I also developed a new approach to infer relations between conceptsin an online encyclopedia.The first and most conspicuous problem I attacked is the classification of online social network(OSN) users into human or automatic accounts (social bots). Starting from a state-of-the-artexisting technology (Botometer) I developed a method that surpasses the main limitation thatplagues many existing approaches based on supervised learning, i.e. their drop in performancewhen tested in domains not considered for training. The idea that inspires my method is theempirical observation that different bot classes are characterized by differences in feature spacethat are found only in class-specific dimensions. My method also exploits the fact that human accounts are more homogeneous compared to bot accounts across domains. The proposed methodcombines the response of specialized classifiers for humans and each class of bots through themaximum rule. This ensemble of specialized classifiers increases the average F1 score from thebaseline of 47% to 73% for unseen accounts. Furthermore, novel bot behaviors are learned withfewer labeled examples during retraining.A second concern of my thesis is the determination of OSN user gender. Also in this case, currentmethods suffer from the problem of poor generalizability to new domains. I developed a supervised learning approach that leverages the tweets and other public information available for theuser account. Using stacked classifiers in conjunction with boosting, I achieved a 96% accuracy that surpasses the current state-of-the-art alternative approaches. I also show that my model basedon portable features can perform well when content is scarce, or when tweets do not include thediscriminating terms that are used by other popular approaches.The last portion of my thesis develops a method that infers prerequisite relations between concepts described in Wikipedia pages by leveraging the navigation patterns of users that visit thosepages. To prove the effectiveness of my method, I introduced (and made publicly available) a newdata set by mapping Metacademy concepts to Wikipedia semantic space. I obtained F1 scores ofaround 80% on the new Metacademy dataset and in the range 82% to 87% for other datasets whichexceeded the state-of-the-art methods.

ISBN: 9798582516637Subjects--Topical Terms:

523869
Computer science.
Subjects--Index Terms:

Classification

Mining User Behaviors in Social Networks Using Machine Learning.
LDR:03669nmm a2200373 4500 001 2281996
005 20210927083514.5
008 220723s2021 ||||||||||||||||| ||eng d
020 $a 9798582516637
035 $a (MiAaPQ)AAI28317657
035 $a AAI28317657
040 $a MiAaPQ $c MiAaPQ
100 1 $a Sayyadiharikandeh, Mohsen. $3 3560718
245 1 0 $a Mining User Behaviors in Social Networks Using Machine Learning.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2021
300 $a 138 p.
500 $a Source: Dissertations Abstracts International, Volume: 82-09, Section: B.
500 $a Advisor: Flammini, Alessandro.
502 $a Thesis (Ph.D.)--Indiana University, 2021.
506 $a This item must not be sold to any third party vendors.
520 $a The work described in my thesis is in the area of applied machine learning. I developed supervisedmethods that leverage user behavior in online social and knowledge networks to infer some specific attributes of such users. I also developed a new approach to infer relations between conceptsin an online encyclopedia.The first and most conspicuous problem I attacked is the classification of online social network(OSN) users into human or automatic accounts (social bots). Starting from a state-of-the-artexisting technology (Botometer) I developed a method that surpasses the main limitation thatplagues many existing approaches based on supervised learning, i.e. their drop in performancewhen tested in domains not considered for training. The idea that inspires my method is theempirical observation that different bot classes are characterized by differences in feature spacethat are found only in class-specific dimensions. My method also exploits the fact that human accounts are more homogeneous compared to bot accounts across domains. The proposed methodcombines the response of specialized classifiers for humans and each class of bots through themaximum rule. This ensemble of specialized classifiers increases the average F1 score from thebaseline of 47% to 73% for unseen accounts. Furthermore, novel bot behaviors are learned withfewer labeled examples during retraining.A second concern of my thesis is the determination of OSN user gender. Also in this case, currentmethods suffer from the problem of poor generalizability to new domains. I developed a supervised learning approach that leverages the tweets and other public information available for theuser account. Using stacked classifiers in conjunction with boosting, I achieved a 96% accuracy that surpasses the current state-of-the-art alternative approaches. I also show that my model basedon portable features can perform well when content is scarce, or when tweets do not include thediscriminating terms that are used by other popular approaches.The last portion of my thesis develops a method that infers prerequisite relations between concepts described in Wikipedia pages by leveraging the navigation patterns of users that visit thosepages. To prove the effectiveness of my method, I introduced (and made publicly available) a newdata set by mapping Metacademy concepts to Wikipedia semantic space. I obtained F1 scores ofaround 80% on the new Metacademy dataset and in the range 82% to 87% for other datasets whichexceeded the state-of-the-art methods.
590 $a School code: 0093.
650 4 $a Computer science. $3 523869
650 4 $a Web studies. $3 2122754
650 4 $a Behavioral sciences. $3 529833
650 4 $a Artificial intelligence. $3 516317
653 $a Classification
653 $a Machine learning
653 $a Social network
653 $a User behaviours
690 $a 0984
690 $a 0602
690 $a 0646
690 $a 0800
710 2 $a Indiana University. $b Computer Science. $3 3560719
773 0 $t Dissertations Abstracts International $g 82-09B.
790 $a 0093
791 $a Ph.D.
792 $a 2021
793 $a English
856 4 0 $u https://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=28317657