Article Preview
Top1. Introduction
In the recent past, social media analytics has gained tremendous momentum due to the vast amount of user-generated data. This data explosion is a gold mine for researchers to perform various analytics to extract insights on different aspects. People vent out their opinions and feelings about diverse topics through social networking sites such as Twitter and Facebook. Globally, Twitter has 330 million monthly active users and 145 million daily active users (Murthy, 2018). There are 22.2 million active members in India. Through Twitter, roughly 500 million tweets are generated per day all over the world and 63% of Twitter users range from 35 to 65 years old. There has been a wide recognition for Twitter which has become a powerful gauge of public sentiments for a spectrum of issues such as social, medical, socio-economic factors on happiness, climate policies, politics, understanding public perception of crime, public health, movie sales, stock market and much more. By generating association rules, taxonomies are built. The main aim of this research work is to extract causal factors from the text-based association rules which are useful in forming a collective opinion on a topic.
Many researchers have used the social media platform to analyze the text posted by the users to obtain the outcomes or insights related to public issues. Association rule mining techniques are used to mine social media data (Chen et al, 2020). Researchers have applied evolutionary computing techniques or nature-inspired optimization techniques to generate meaningful association rules. However, only a few have experimented with the clustering of text-based association rules. A hybrid Firefly – Ontology-based clustering algorithm (FF-OC) is proposed to cluster the association rules. The clusters thus formed provide an insight into the causal factors affecting a topic discussed widely in social media. The methodology uses a keyword-based ontology to compute the similarity among features (Zamazal, 2020). Firefly based algorithmic technique is one of the most popular meta-heuristic based algorithms developed by Xin-She Yang (Yang, 2008) based on the behavior of real fireflies. The algorithm is efficiently applied in many domains such as text clustering, image processing, data analytics, and classification. The advantage of the firefly algorithm over the existing swarm intelligence-based algorithms is that it converges quickly (Xie et al., 2019). Since the algorithm automatically split its population into subgroups, each subgroup can find the best global solution.
The following enumerates the contributions of the proposed work:
- 1.
Generating association rules from the text corpus collected from Twitter is a challenging task. Generally ARs reflect an association or context between items in the left hand side and right hand side of the rule. In this work, WordNet is used to incorporate such context for each of the association rules generated from the text features extracted from the twitter data.
- 2.
The algorithm generates a huge number of ARs. Many of them do not contribute to the effectiveness of the accuracy of the results. Therefore an algorithm is designed to prune the repetitive ARs.
- 3.
Though many ARs are pruned, since the collected text corpus is large, the resultant set is still huge. An evolutionary, nature-inspired algorithm is designed for clustering the ARs by performing an exhaustive exploration of all the ARs.
- 4.
The main contribution and novelty lie in the fact that no prior knowledge is applied during the clustering process. From the clusters thus generated, causal factors are identified.
The algorithm that is proposed in this paper differs from the existing works in literature in the way that the association rules are clustered without any prior knowledge. Further insights are provided as impacts or consequences. The theoretical implication of this work is that a new model for clustering association rules is formulated. The clusters that are obtained from the model give insights into various factors that affect the domain of interest. For instance, the government may take action on implementing certain regulations based on the factors impacting which are the main outputs from the research. The main research objectives are listed as follows: