Identifying Potential Customer Using Graph Social Media Analytics

In today's digital era, businesses try to find new clients in many ways. This study offers approaches that use graph social media analytics to identify potential clients for certain products. In this study, we consider Soft Pinch Liquid Blush. The dataset used was gathered from Twitter search queries. The main objective of this study is to discover potential customers for the product Soft Pinch Liquid Blush using graph social media analytics, as well as find significant nodes relevant to the product. The analysis starts by building a graph network with the Source and Target columns, where the Source column represents people who tweeted about the product and the Target column represents people who were referenced or reacted to a tweet, indicating prospective interest. To evaluate the influence nodes, voterank is used. Then, to find communities in the network, three graph community techniques which are Louvain, Clauset-Newman-Moore (CNM) greedy modularity and label propagation (LP) are used. The community results were compared using modularity, coverage and performance. Based on the result, node "rarebeauty" was the most influential node based on voterank. Then, Louvain algorithm detected 1265 communities, LP produced 1412 communities and CNM produced 1318 communities after applying the graph algorithms. In terms of modularity and performance scores of community structure, the Louvain and CNM algorithms appear to outperform LP algorithm. However, LP algorithm outperforms Louvain and CNM in terms of coverage. The study concludes that Louvain and CNM are the suitable algorithms in this study, even though the modularity value was low. For future research, it recommended to add more edges between nodes to increase the overall connection of the network and do network pre-processing such as noise removal, sparsity reduction and edge weight normalization to enhance the modularity score.