Social Media Encourages Public Transport Analysis in the Maldives
Famous for its many high rises and urban neighbourhoods, Malé (the capital of the Maldives) is one of the most densely populated cities in the world. In the late 1990s, the Government began a reclamation project to develop the nearby island of Hulhumalé in order to accommodate the capital’s growing population.
Joining forces with our colleagues at UNDP Maldives and UNDP Bangkok Regional Hub, Pulse Lab Jakarta designed a pair of research projects to investigate urban mobility patterns using origin-destination analysis, and to perform sentiment analysis of public transport using Twitter data in order to better understand people’s commuting experiences from Hulhumalé. To help us achieve these goals, we also consulted with several local players in the Maldives, including the National Bureau of Statistics, Communications Authority of Maldives, Maldives Transport and Construction Company, Ooredoo Maldives and the state-owned Housing Development Corporation.
For residents of Hulhumalé, many of the concerns regarding planning and the provision of public services revolve around issues of transport, such as service irregularity and underdeveloped infrastructure. Recent state policies, however, have resulted in the establishment of new targets for the city, which have underlined the need for exploring the residents’ current experiences to inform future development decisions.
Accordingly, the overall objectives of this research are multifaceted. At a macro level, sentiment analysis is intended to provide a general understanding about citizen perceptions of public transportation, accentuating any major issues that may be worth examining. With origin-destination analysis, on the other hand, the expectation is to provide evidence to policy makers in order to facilitate decision making at all levels.
Inspired by one of the Lab’s past research projects in which locational information from social media on mobile devices was successfully used to reveal commuting patterns in the Greater Jakarta area, we had similarly hoped to examine GPS-stamped tweets in Hulhumalé to conduct the origin-destination analysis. Unfortunately, we were unable to use this approach due to a lack of geo-tagged tweets (while many residents on the island are active on Twitter, a relatively small number uses the geo-tagging feature). So as a first step, our research team performed the research using sentiment analysis.
Below we share some preliminary results.
Identifying The Focus Areas
Following consultations with one of our partners on the ground, the urban planning and community engagement team at the Housing Development Corporation, we identified four areas to focus on: ferry transportation, bus transportation, taxi transportation, and parking. To capture relevant citizen feedback shared via Twitter, our research team used a set of semi-automatically-defined keywords, whereby initial keywords were selected based on discussions with domain experts, followed by an interactive optimisation approach to identify broad terms and then narrow them to enhance certain keywords.
In addition to understanding the various sentiments associated with each topic, the research team conducted: volume trends analysis (to understand how the message dynamically changes over time); topic and keyword breakdown analysis (to understand the detail of conversation of selected topic); account analysis (to examine players who are actively engaged or frequently tagged in content related to a certain topic); and highlights (to gain more detailed insights).
Visualising The Results
Based on the amalgamated citizen feedback via Twitter on all four focus areas, i.e. ferry transportation, bus transportation, taxi transportation, and parking, our findings revealed that there were considerably more negative than positive sentiments expressed, 41 per cent to 15 per cent respectively. Nonetheless, based on the data collected during the period under examination, the majority of the sentiments shared about public transport were neutral, i.e. not explicitly negative or positive (44 per cent).
Additionally, the volume of tweets related to transportation was observed to be dynamically changing, with some noticeable increases that were hypothesised to be caused by related headline news during those periods. For instance, towards the end of May 2016, the volume of tweets surged, which coincided with an announcement in the local news about the provision of motorcycle taxi services in the Maldives.
Combined with additional research (such as analysis of wifi hotspots data and call detail records to capture insights on mobility), these results will be useful for informing future public transportation policies and other development approaches. The sentiment analysis conducted, though limited to Twitter data, also demonstrates the potential utility of using wider social media data to derive insights on different aspects of public transport, especially considering the island’s growing population (from 1,000 residents in 2004 to more than 40,000 as of 2016).
At the moment, the research team is conducting further analysis and we’ll share the results as they become available. The Lab was delighted for the invitation to collaborate on this project, and is keen on establishing other partnerships to work on projects for social good.
Pulse Lab Jakarta is grateful for the generous support from the Government of Australia.