Academia.eduAcademia.edu

Figure 5 – uploaded by Zuraini Zainol

See full PDFdownloadDownload figure
Fig. 5: A screenshot for item frequency histogram In this section, the experimental results for the proposed frequent keyword patterns discovery in cyberbully tweets are discussed. 8275 tweets were used as an input dataset for frequent pattern mining. In this experiment, 16 keywords represent as items while 8275 tweets are transactions. The frequencies of each item are presented in Figure 5. Figure 5 presents the item frequency histogram that illustrates how many times an item has occurred in the dataset as compared to the others. The items ‘sial’(3244), ‘bodoh’(2755) and ‘babi’(1329) had the highest frequencies in the 8275 Twitter data. This indi- cates that, these items are frequently used keywords by most Ma- laysian tweeters to bully someone on Twitter.

Figure 5 A screenshot for item frequency histogram In this section, the experimental results for the proposed frequent keyword patterns discovery in cyberbully tweets are discussed. 8275 tweets were used as an input dataset for frequent pattern mining. In this experiment, 16 keywords represent as items while 8275 tweets are transactions. The frequencies of each item are presented in Figure 5. Figure 5 presents the item frequency histogram that illustrates how many times an item has occurred in the dataset as compared to the others. The items ‘sial’(3244), ‘bodoh’(2755) and ‘babi’(1329) had the highest frequencies in the 8275 Twitter data. This indi- cates that, these items are frequently used keywords by most Ma- laysian tweeters to bully someone on Twitter.

Related Figures (9)

Copyright © 2018 Authors. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. The remainder of the paper is organized as follows: Section 2 explores related work and discusses different methods that have been used for cyberbullying detection. Section 3 discusses the experiments details, where Apriori and Association Rule Mining are used to analyse the research problem. Section 4 presents the experimental results. Finally, the conclusion and future work of the study are made in Section 5.
Copyright © 2018 Authors. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. The remainder of the paper is organized as follows: Section 2 explores related work and discusses different methods that have been used for cyberbullying detection. Section 3 discusses the experiments details, where Apriori and Association Rule Mining are used to analyse the research problem. Section 4 presents the experimental results. Finally, the conclusion and future work of the study are made in Section 5.
Fig. 1: The Framework for Discovering Frequent Keyword Patterns and Relationship (DFKPR) adopted from [25] Singh et. al. [22] advocates using fusion approaches to combine heterogeneous sources for cyberbullying detection. The approach In previous study [25], we proposed a framework for discovering interesting frequent patterns and relationships using Apriori Algo- rithm. Also, the algorithm has been discussed in this work. The framework consists of three main stages: (i) data collection and pre-processing, (ii) frequent keyword pattern identification, and (iii) interesting relationships discovery. This framework applied frequent keyword pattern mining and association rules mining for extracting interesting relationships between keyword and the Had- ith Chapters from the Book of Friday prayer. By using the tech- nique of Association Rule Mining (ARM), frequent patterns and interesting relationships that can benefits Muslim scholars in mak- ing full use of Hadith in assisting them in their daily Islamic life and practice. ARM has been widely applied in many application domains such as healthcare [26], predicting flood areas [27], text mining [25], education, etc. In this study, we implement the framework of [25] over the tweets dataset on cyberbullying (see Figure 1).
Fig. 1: The Framework for Discovering Frequent Keyword Patterns and Relationship (DFKPR) adopted from [25] Singh et. al. [22] advocates using fusion approaches to combine heterogeneous sources for cyberbullying detection. The approach In previous study [25], we proposed a framework for discovering interesting frequent patterns and relationships using Apriori Algo- rithm. Also, the algorithm has been discussed in this work. The framework consists of three main stages: (i) data collection and pre-processing, (ii) frequent keyword pattern identification, and (iii) interesting relationships discovery. This framework applied frequent keyword pattern mining and association rules mining for extracting interesting relationships between keyword and the Had- ith Chapters from the Book of Friday prayer. By using the tech- nique of Association Rule Mining (ARM), frequent patterns and interesting relationships that can benefits Muslim scholars in mak- ing full use of Hadith in assisting them in their daily Islamic life and practice. ARM has been widely applied in many application domains such as healthcare [26], predicting flood areas [27], text mining [25], education, etc. In this study, we implement the framework of [25] over the tweets dataset on cyberbullying (see Figure 1).
Fig. 2: A screenshot for extracting text data in RapidMiner This section provides the results from the experiment carried out using DFKPR framework. Data collection is a crucial phase in this research. Researchers spent a lot of time on collecting text data from Twitter and preprocessing task. In this study, we collected text data from Twitter using RapidMiner 7.5.3. RapidMiner is a powerful visual workflow designer for building predictive analytic workflows. RapidMiner also offers hundreds features of data preparation (e.g., sampling, transformations, weighting and selec- tion, attribute generation, etc.), machine learning algorithms to support data science projects [28]. In this study, we applied the Twitter Connector for accessing the Twitter data (phrases, tweets or user profile information) directly from RapidMiner Studio. The ‘Search Twitter’ operator (available by the RapidMiner “Social Media” extension) enables users to search the top 1000 most re- cent posting tweets, provided that a Twitter connection has been configured first in this experiment.
Fig. 2: A screenshot for extracting text data in RapidMiner This section provides the results from the experiment carried out using DFKPR framework. Data collection is a crucial phase in this research. Researchers spent a lot of time on collecting text data from Twitter and preprocessing task. In this study, we collected text data from Twitter using RapidMiner 7.5.3. RapidMiner is a powerful visual workflow designer for building predictive analytic workflows. RapidMiner also offers hundreds features of data preparation (e.g., sampling, transformations, weighting and selec- tion, attribute generation, etc.), machine learning algorithms to support data science projects [28]. In this study, we applied the Twitter Connector for accessing the Twitter data (phrases, tweets or user profile information) directly from RapidMiner Studio. The ‘Search Twitter’ operator (available by the RapidMiner “Social Media” extension) enables users to search the top 1000 most re- cent posting tweets, provided that a Twitter connection has been configured first in this experiment.
Fig. 4: A screenshot for setting the tweets details that contain specific keywords We also filtered our search by removing retweets (-rt) and links (- http) (see Figure 3). The result containing all posting tweets text containing the specified keywords or terms with their tweets metadata will be displayed in the RapidMiner format (see Figure 4).
Fig. 4: A screenshot for setting the tweets details that contain specific keywords We also filtered our search by removing retweets (-rt) and links (- http) (see Figure 3). The result containing all posting tweets text containing the specified keywords or terms with their tweets metadata will be displayed in the RapidMiner format (see Figure 4).
Fig. 3: A screenshot for setting the tweets details that contain specific keywords Figure 2 shows the implementation of three operators in the pro- cess of collecting Twitter data. In the “Search Twitter” operator, users can set up the requirement of tweets by filling up the given parameters (see Figure 3). For example, the number of tweets to be returned is set up to 1000 and only recent tweets need to be returned to RapidMiner Studio is specified in the “Result type” parameter.
Fig. 3: A screenshot for setting the tweets details that contain specific keywords Figure 2 shows the implementation of three operators in the pro- cess of collecting Twitter data. In the “Search Twitter” operator, users can set up the requirement of tweets by filling up the given parameters (see Figure 3). For example, the number of tweets to be returned is set up to 1000 and only recent tweets need to be returned to RapidMiner Studio is specified in the “Result type” parameter.
Fig. 6: The result after generating Association Rules based on ‘support’ measurement
Fig. 6: The result after generating Association Rules based on ‘support’ measurement
Fig. 8: Visualization of network graph for 10 association rules according to lift Among the 88 rules are produced in the experiment, rules 1 and 2 have the highest lift at 30.49 (see Figure 7). This shows that whenever Malaysian tweeters posted the keywords ‘pendek’, they are very likely to post the keyword ‘gemuk’. This result can also be visualized in form of an interactive plot in Figure 9.
Fig. 8: Visualization of network graph for 10 association rules according to lift Among the 88 rules are produced in the experiment, rules 1 and 2 have the highest lift at 30.49 (see Figure 7). This shows that whenever Malaysian tweeters posted the keywords ‘pendek’, they are very likely to post the keyword ‘gemuk’. This result can also be visualized in form of an interactive plot in Figure 9.
Fig. 9: An interactive plot for 88 association rules
Fig. 9: An interactive plot for 88 association rules
Related topics:
Association Rules MiningCyberbullying Detection
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved