Incorporating crowdsourced social media footprint in public transport service quality assessment framework

Abstract

Public participation holds a pivotal value in any decision-making. But conventional public participation inherits specific limitations. With internet 2.0, epublic participation (technology-mediated, facilitated by ICT tools) has emerged as a plausible adjunct to the limitations that conventional participation possesses. In this regard, crowdsourced data has been a profound gain as it seeks to engage the public using the internet, smartphone apps, or social media. Researchers have assessed the potential of crowdsourced data in transport, health, and environmental issues. This research focuses on using crowdsourced data for transport planning since it is dynamic, and so social media data can offer compelling support. Also, envisioning our cities, we have coined the term “smart cities” and “smart mobility” is a strategic factor of it. The emphasis is on encouraging people to use public transport than owning a car. As a result, the research narrowed its focus to public transport. The service quality assessment (SQA) aspires to improve the efficiency of the existing public transport system. The users’ perception of the qualitative aspect of service is of utmost importance for improving the quality of service. But SQA frameworks have a limited deliberation for it. In India, the MoUD (now MoHUA) service benchmark for public transport comprises six parameters, but only two related to the qualitative aspect. Even these two, however, do not consider the users’ perception. This brings an opportunity to incorporate crowdsourced social media data to integrate the qualitative aspect into the SQA framework. At the moment, there has been a minimal study on incorporating social media footprint into the existing SQA. And this research would add a new value to the field of study. The research considers the Delhi metro as a pilot case, a member of the Community of Metro Group (CoMET), which conducts the customer satisfaction survey for Delhi metro. The proposition is that this approach is an addition to the existing structure that will assist to update feedback data. The data collection for this research is two-fold. First, tweets from the official handle of the Delhi metro are extracted for a month. Second, for ground truth and to bring a structure to the feedback data, tweets in a separate Twitter handle made for the research purpose have been extracted. Throughout this research, Google Colaboratory, a cloud-based Jupyter notebook environment, is used. After data pre-processing, topics are predicted from the extracted data using Mallet’s LDA machine learning model. The BERT cased machine learning model is used to predict the sentiments of each of the identified topics. As the research concerns feedback analysis, we train the model on reviews that are tagged on a scale of one to five and are publicly available. The research performs factor analysis for two reasons: first, to compute the final topics (or sub-dimensions) using the factor loading method, and second, to classify the identified topics under common factors (or primary dimensions). The research further adopts the weighted SERVPERF scale, for calculating the service quality of each sub-dimension, primary dimension, and overall structure. Finally, the research employs the Importance-Performance analysis to single out the areas where improvement will have the greatest effect on improving overall system performance. The result highlights 11 topic (or sub-dimensions) classified under four factors (or primary dimensions) for which the perceived service quality is measured. It pinpoints the precise niche within each topic (or sub-dimension) that is important to users. Also, a prototype framework for incorporating social media analytics in the service quality assessment is worked out. The research’s key finding is that the novel approach encourages users to share their service experiences based on their comfort and, from the provider’s perspective, dynamic insight into how users perceive the service.