| dc.description.abstract | The purpose of this study was to establish the functional loopholes in the customer support rendered in the telecom industry in Uganda, with a major focus on Twitter as the support channel, in order to assess how best the processes therein can be automated using machine learning in order to improve the overall customer experience. In this research, we considered a descriptive research design, and an endeavour was taken to use both qualitative and quantitative data to establish the state of customer support in the telecom industry in Uganda, as well as how best machine learning can be leveraged to improve it through automation. The target population consisted of 155 volunteers and the two biggest telecom service providers in Uganda by market share, that is MTN and AIRTEL. Random sampling was used to select the online survey respondents, who were majorly students at Makerere University. 1003 customer support tweets sent to MTN and AIRTEL before 15th September 2022 were programmatically collected and analyzed to assess the the performance of these companies across various metrics in customer support, as well as to validate and justify our proposed method of automation. An online survey and the Twitter API were employed during the collection of data. Microsoft Excel was then extensively used to clean, code and enrich this data. Python was the primary language and tool used in the analysis of the data. The Numpy and Pandas libraries were used in the exploratory analysis of the data and Matplotlib was used in the explanatory, or rather visualization phase of the data analysis.
When correlations were required, the Pearson correlation coefficient was used. After extensive research through existing literature, sentiment analysis using machine learning was proposed as a solution to automate some processes in the pipeline of a trouble ticket system. Two libraries, that is — PyABSA and TextBlob were used to analyze and justify this proposal. The research findings to a large degree proved the necessity for handling customer support in a platform or environment dedicated
to that purpose, since such environments offer the best opportunities to automate processes in trouble
ticket systems. The research paper concludes with a few recommendations on how best this research can be used to not
only support subsequent research in this area, but also guide stakeholders concerned with customer care,
support and experience across various domains. Specifically, further research is encouraged in the automation of processes involved in e-governance, since it has one of the biggest amounts of data and highest
need for automation. | en_US |
