Quantify how cost, time, distance, comfort, and income influence inter-district mode choice among BUET resident students traveling from Dhaka, and develop a validated model that supports policy scenarios on fares, service reliability, and comfort upgrades.
Discrete choice modeling in R using Apollo; the core specification is Multinomial Logit with alternative-specific utility terms.
Thirteen ascending specifications from ASC only to full attribute sets and interactions; selection guided by log likelihood, AIC, BIC, adjusted rho squared, and t statistics.
Variable scaling and normalization for numerical stability; explicit availability vectors per respondent and per alternative.
Goodness of fit, sign and magnitude checks, and reasonableness tests; sensitivity runs on key attributes.
Revealed preference survey of 241 BUET resident students; origin Dhaka to home district.
Instrument design, piloting, and revision; capture of socioeconomics, vehicle ownership, trip purpose, generalized cost, comfort, and stated availability.
Cleaning and coding pipeline in R: outlier screening, missingness handling, normalization of continuous attributes, and construction of alternative specific design matrices.
Progressive ladder of 13 models; each step adds interpretable structure and is retained only if information criteria improve.
Final model achieves adjusted rho-squared near 0.47 with coherent signs on all retained coefficients.
Robustness checks: attribute rescaling, removal of rare alternatives during estimation, stability of key coefficients under small perturbations.
Documentation of IIA awareness and future extension path to nested or mixed logit if a public versus private nesting becomes necessary.
Price sensitivity is strongest for the bus and the AC bus; a higher fare reduces choice probability as expected.
Travel time carries negative utility for AC bus, train, and rental; personal car time is comparatively weak in this student cohort.
Distance increases the appeal of AC bus and train and reduces the appeal of rental, consistent with long-haul versus short range behavior.
Family income loads on personal car preference; self income supports rental usage.
Comfort improves the utility of AC bus and personal car and is positive for rental, reflecting quality seeking on longer trips.
Student fare instruments will shift bus and AC bus shares; price elastic segments can be targeted without eroding total throughput.
For longer distances, reliability and comfort enhancements on AC bus and train have a clear payoff; priority should be punctuality and seating quality.
Rental behaves as a short range substitute; design safe, well-signed pick-up and drop off zones near campus and major terminals.
Segment communications by income and car access; personal car-owning households require a different message than self funded renters.
Bus and AC bus are price-sensitive within this cohort; cost reductions convert quickly to share gains.
Time penalties matter for AC bus, train, and rental; reliability is as salient as speed for long trips.
Distance steers riders toward AC bus and train and away from rental, which functions as a short range solution.
Family income aligns with personal car use; self-income maps to rental.
Comfort raises utility most for AC bus and the personal car.
The estimated structure is suitable for controlling what-if analysis on fare and time policies in student travel markets.
Service planners should pair fare instruments with targeted reliability and comfort upgrades on long-distance services.
A mixed or nested extension can capture taste heterogeneity and relax IIA for finer policy design without discarding the current pipeline.