Enhancing Clustering of Users in Social Media Networks for Improved Digital Marketing

Abstract

The Internet has changed the world in which we sell. It reaches beyond being a new channel for marketing and offers a new paradigm for the way consumers connect with brands and with each other. Online Social Networks (OSN) which began in the form of generalized online communities that focused on bringing people together to interact with each other have now become an avenue for marketers to look for customers. The rapidly expanding social network audiences in the emerging markets will be huge drivers of social user growth. Changes affecting traditional marketing have been seen where people now spend less time watching TV and reading print newspapers each day and instead communicate through the use of mobile phones, watch videos on YouTube, read the newspaper online, look at photos in Flickr and exchange information through social networks. This transition has forced businesses to find alternative affordable ways to reach customers. The big question for marketers is how to convert users of OSN to customers. The purpose of this research work was Enhancing Clustering of Users in Social Media Networks for Improved Digital Marketing. This is in recognition of the fact that many customers have social media network accounts such as twitter and facebook that can be effectively utilized to convey advertisement information to them. Using these platforms, the digital marketers have access to behavioural and demographic data that can be mined to extract actionable patterns. However, social media data can contain a large portion of noisy data. As such, the definition of noise becomes complicated and relative because it is dependent on the task at hand. Therefore, this study sought to address the challenges of noise removal from social media data and clustering users in social media networks for improved digital marketing. The researcher formulated five objectives which were to: Identify noise in data depending on task at hand for enhanced clustering; Use weight allocation in Term Frequency-Inverse Document Frequency (TF-IDF) environment in order to enhance detection of noise in data; Determine optimal model for reducing noise in data; Make deductions that will lead to design of appropriate framework for social media networks marketing; and to test the clustering framework for social media network adverts. To achieve these objectives, xvi a quantitative research design was adopted. The data was collected from Facebook social media network open groups, namely Soko Kuu and Soko Nyeusi Official Group and coded into numerical format to allow for mathematical analysis. This study delimited itself to user demographics data that may influence customer buying trends. A target sample of 155 was employed and TF-IDF values were employed as a basis for noise elimination. The study reveals social media data is indeed noisy and identification of the noise from the data depended on task at hand. Uniquely identifiable Keywords were formulated from the data and tf-idf weighting computed on these keywords as a basis for further eliminating of noise. Keywords that appeared in nearly all documents were given a low weight. Noise identification was achieved by observing the tf-idf values for the different measurable constructs and words with very low tf-idf values compared to their counterparts were regarded as noise and eliminated from further analysis. For the dataset used in this study Nationality, Youth, Group membership and Market were noisy and eliminated from further analysis. It was noted for latent behavioral variables elimination of variables with low tf-idf values as noise is not always obvious as constructs with lower tf-idf values may be indications for some interesting insights. To determine the optimal model for reducing noise in data, variables were adopted or dropped depending on their path costs. Results obtained from the model revealed that all the measurable variables were significantly above the threshold value of 0.05 implying the developed model was the attuned one for the given dataset. Indicators were hypothesized consisting of both behavioral and demographics variables to make deductions leading to design of appropriate framework for social media marketing. In this study demographic variables were used to design appropriate framework for social media marketing. Where tf-idf values for a comparable variable were higher than the other, advertisements were skewed to fit the variable with higher tf-idf value. In this study for automobile goods and services category, tf-idf values for adults were much higher than that of the youth; the consequence is that advertisements should be more skewed to fit adults than youth. Testing the clustering framework for social media marketing revealed among all the xvii variables buyer attitudes were the most important while religious views were the least important when arranged in their order of importance in the clusters formed. Hypothesis was formulated and were accepted or dropped from the results obtained. Possible combinations between demographics and behavioral constructs were carried out to test the hypothesis. When the test was run hypothesis whose significance levels were above the threshold of 0.05 had they null hypothesis accepted meaning the variables influenced each other. In the case that had null hypothesis rejected, the alternate hypothesis stated that the two constructs never influenced each other. In this study from the dataset used, the deductions made were physical address influenced buyer readiness, buyer attitude and benefit sought. Gender did not influence buyer readiness, buyer attitude or benefit sought. Age influenced buyer readiness and buyer attitude but did not influence benefit sought. Education did not influence buyer readiness, buyer attitude or benefit sought. All these contextual issues informed the study on enhancing clustering of users in social media networks for improved digital marketing. A set of recommendations and guidelines is presented, which could act as a refence point for improving digital marketing in social media networks not only in Kenya, but also in the wider world.