Mining the Attitude of Social Network Users using K-means Clustering

Gurusamy, Vairaprakash; Kannan, S.; Jayagopal, Prabhu

doi:10.23956/ijarcsse/sv7i5/0231

Cited by 8 publications

(6 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Both of these models are token-level based. Tokens are small units (e.g., words, phrases, symbols or other meaningful elements Gurusamy and Kannan 2015) into which a text is split.…”

Section: Methodsmentioning

confidence: 99%

“…We lowercased each post to avoid word types such as 'Hello' and 'hello' being treated differently (Pradha, Halgamuge, and Tran Quoc Vinh 2019). We also removed punctuation, URLs, @ mentions and hashtags because they may not provide much information to the text and may be noisy (Gurusamy and Kannan 2015). Finally, we used text embeddings as input features which are generated using the Universal Sentence Encoder (Cer et al 2018).…”

Section: Data Pre-processingmentioning

confidence: 99%

See 1 more Smart Citation

Automated hate speech detection and span extraction in underground hacking and extremist forums

et al. 2022

View full text Add to dashboard Cite

Hate speech is any kind of communication that attacks a person or a group based on their characteristics, such as gender, religion and race. Due to the availability of online platforms where people can express their (hateful) opinions, the amount of hate speech is steadily increasing that often leads to offline hate crimes. This paper focuses on understanding and detecting hate speech in underground hacking and extremist forums where cybercriminals and extremists, respectively, communicate with each other, and some of them are associated with criminal activity. Moreover, due to the lengthy posts, it would be beneficial to identify the specific span of text containing hateful content in order to assist site moderators with the removal of hate speech. This paper describes a hate speech dataset composed of posts extracted from HackForums, an online hacking forum, and Stormfront and Incels.co, two extremist forums. We combined our dataset with a Twitter hate speech dataset to train a multi-platform classifier. Our evaluation shows that a classifier trained on multiple sources of data does not always improve the performance compared to a mono-platform classifier. Finally, this is the first work on extracting hate speech spans from longer texts. The paper fine-tunes BERT (Bidirectional Encoder Representations from Transformers) and adopts two approaches – span prediction and sequence labelling. Both approaches successfully extract hateful spans and achieve an F1-score of at least 69%.

show abstract

“…Both of these models are token-level based. Tokens are small units (e.g., words, phrases, symbols or other meaningful elements Gurusamy and Kannan 2015) into which a text is split.…”

Section: Methodsmentioning

confidence: 99%

Section: Data Pre-processingmentioning

confidence: 99%

Automated hate speech detection and span extraction in underground hacking and extremist forums

et al. 2022

View full text Add to dashboard Cite

show abstract

“…The relationship of between its datasets and the center of its cluster is the shortest distance. This algorithm has been used to cluster the earthquake epicenter in [4], to mining the attitude of social network users in [5], and mapping of image and video in [6]. The uniform effects of this algorithm have studied in [7].…”

Section: Introductionmentioning

confidence: 99%

A Performance of the Scattered Averaging Technique based on the Dataset for the Cluster Center Initialization

Putra¹,

Gaffar²,

Suprapty³

et al. 2021

IJMECS

View full text Add to dashboard Cite

Clustering is one of the primary functions in data mining explorations and statistical data analysis which widely used in various fields. There are two types of the clustering algorithms which try to optimize certain objective function, i.e. the hierarchical and partitional clustering. This study focuses on the achievement of the best cluster results of the hard and soft clustering (K-Mean, FCM, and SOM clustering). The validation index called GOS (Global Optimum Solution) used to evaluate the cluster results. GOS index defined as a ratio of the distance variance within a cluster to the distance variance between clusters. The aim of this study is to produce the best GOS index through the use of the proposed method called the scattered averaging technique based on datasets for the cluster center initialization. The cluster results of each algorithm are also compared to determine the best GOS index between them. By using the annual rainfall data as the dataset, the results of this study showed that the proposed method significantly improved K-Mean clustering ability to achieve the global optimum solution with a performance ratio of 69.05% of the total performance of the three algorithms. The next best clustering algorithm is SOM clustering (24.65%) followed by FCM clustering (6.30%). In addition, the results of this study also showed that the three clustering algorithms achieve their best global optimum solution at the number of even clusters.

show abstract

“…K-Means Clustering is a new method of partition that has mutually exclusive clusters that are spherical in shape. It can generate a certain number of flat (nonhierarchical) and disjoint clusters that organize the objects into k -partitions wherein every partition is a cluster [5].…”

Section: Introductionmentioning

confidence: 99%

Data Leak Identification in Social Networks Using Hybrid Tranquility K Means Clustering

Karthik¹,

Tamizhazhagan²,

Narayana³

2021

INDJCSE

View full text Add to dashboard Cite

Loss of sensitive data can be stopped employing Data Leak Prevention (DLP). Most of such tools can be quite effective while protecting private information known already. At the same time, plenty of private information has not been recognized until it has been disclosed to various unknown users or other competition enterprises. Clustering refers to a data mining technique that can classify a certain set of instances into different clustersusing a measure of similarity. One of the most common algorithms based on partitioning is the K-Means. However, it has many drawbacks like it can generate local optimal solutions that are based on initial centroids that are chosen randomly. The Tabu Search (TS) Tranquility Search and the Stochastic Diffusion Search (SDS) have been proposed in this work. In one of the most recent algorithms called Tranquility Search, optimal global solutions are obtained by exploring through the entire solution space. Some studies show hybrid algorithms that are a combination of two different ideas producing better solutions. In this work, a new approach is presented which is a combination of two different ideas producing better solutions. The Improved Tranquility Search technique and the K-means algorithm are combined. For this, a hybrid Tranquility-TABU-SDS algorithm is applied in the social network for the DLP. The results of the experiment have proved that the method proposed performs better in comparison to other methods.

show abstract

Mining the Attitude of Social Network Users using K-means Clustering

Cited by 8 publications

References 9 publications

Automated hate speech detection and span extraction in underground hacking and extremist forums

Automated hate speech detection and span extraction in underground hacking and extremist forums

A Performance of the Scattered Averaging Technique based on the Dataset for the Cluster Center Initialization

Data Leak Identification in Social Networks Using Hybrid Tranquility K Means Clustering

Contact Info

Product

Resources

About