Investigating road user's perception of urban road traffic congestion: A survey in Mogadishu, Somalia

Cities often experience traffic congestion, which can negatively affect the local economy, environment and standard of living. Urban congestion is a critical issue affecting cities worldwide, but its impacts on post-conflict urban areas have been understudied. Somalia's capital, Mogadishu, faces severe traffic congestion due to decades of conflict and lack of adequate urban planning. Existing literature on traffic congestion predominantly focuses on stable areas, leaving a significant gap in understanding the dynamics of traffic congestion in conflict-affected cities. This research aims to fill this gap by identifying key factors contributing to traffic congestion from the perspective of road users, assessing the socio-economic impacts of these factors, developing predictive models using random forest machine learning techniques, and proposing sustainable solutions suitable for Mogadishu's unique urban structure. The study focuses on Mogadishu, a city of about 2.5 million inhabitants experiencing rapid urbanization amid inadequate transport systems. The city's rapid urbanization, coupled with poorly maintained roads, inadequate traffic signalization systems, and inadequate public transport infrastructure have exacerbated traffic congestion problems. To achieve the objectives of the study, a questionnaire was administered to 500 respondents, primarily university students, during the data collection process. The questionnaire was designed to target variables such as sociodemographic characteristics, transportation behavior, and perceptions on the causes and solutions to traffic congestion. The research uses Chi-Square test to explore the relationships between various traffic related factors and Delay The study developed four Random Forest models to assess the effects on delay. The first model achieved 93% accuracy, identifying natural factors (weather conditions), inadequate infrastructure and drivers' non-compliance with traffic rules as the most critical factors. The second model evaluated the effectiveness of the proposed solutions, achieving 87% accuracy, and identified the most effective measures as implementing traffic education programs, commercial regulations, removing safety checkpoints, reducing the number of bajaj (three-wheelers) and improving road infrastructure. The third and fourth models achieved 96% and 89% accuracy, respectively, using the scores obtained from factor analysis, showing that grouping variables improves model performance. The results show that reducing complexity through factor analysis improves prediction accuracy and contributes to better traffic management decisions, while precision, recall, and F1 score measures confirm the reliability of the model's predictions. This research fills an important gap in the literature by addressing traffic congestion in a conflictaffected city. The combination of statistical methods and machine learning techniques provides a comprehensive framework for understanding and addressing traffic congestion. Policy makers and urban planners can leverage these findings to develop effective interventions, improve mobility and reduce the socio-economic burdens of traffic congestion in Mogadishu. Moreover, the study's insights can be applied to other cities with similar post-conflict or underdeveloped transportation systems, providing a replicable model for addressing urban mobility challenges.