KNUST Research Week and Scientific Conference: Empowering Researchers to Achieve Impactful Outcomes

In contrast to most industrialized countries, the majority of Sub-Saharan African cities continue to struggle with the growing demand for public transport (PT). Freetown, Republic of Sierra Leone, is exploring proposals to implement the first dedicated bus lane PT system (DBLPT) with high-quality buses on selected corridors. This study leverages AI-based and statistical modelling to calibrate a discrete choice model for conventional paratransit services. It further explored user preference for the planned DBLPT system, its benefits in terms of revenue and passenger ridership and quantified the extent to which the AI-based models are sensitive to changes in transport policies. The findings revealed a strong preference for minibuses (52.0%) and three-wheelers (32.6%), followed by paratransit buses (15.4%) as traditional PT modes. The “Bus and Minibus” system was predicted as the most preferred DBLPT system (54.8%), followed by the “Bus and Three-wheeler” (22.7%). Additionally, the “Bus and Minibus” system was predicted to generate the highest daily revenue and passenger ridership, followed by the “Bus and Three-wheeler” system. The AI-based models produced comparable results that outperform the traditional MNL model. The MNL model exhibits an average prediction accuracy of 80. 9%, while that of the AI-based models ranges from 90% to 94.4%. The latter models were found to significantly improve the prediction accuracy of the calibrated mode choice model by approximately 14%, suggesting the effectiveness of this approach in travel behaviour prediction. Travel cost (TC) was revealed as the most important trip-related attribute, followed by bus stop waiting time (BS_WT) and region of residence. The models are sensitive to TC and BS_WT, suggesting their significant role in commuter mode choice. These models are thus policy sensitive and remain useful for accurately forecasting demand across different modes. These findings and other model results underscore the preference for and benefits of an integrated DBLPT system.