APPLYING BAYESIAN DYNAMIC MIXED LOGISTIC REGRESSION TO MOBILITY NETWORKS

This study explored the performance of Bayesian Dynamic Mixed Logistic Regression Model (BDML) with different priors that include; Beta, Gamma, Cauchy, Exponential, Normal, Jeffrey and Uniform prior. The primary objective of the model was to capture time-varying random intercepts and slopes while accommodating dynamic data structure. The major aim of this research was to compare the BDML model with alternative models including the Bayesian Mixed Logit model, mixed logit and logistic regression and to evaluate their performance. Simulated transportation data revealed that the DBML model outperformed other models; with the modified Bayesian Dynamic Mixed Logit (BDML) model achieving the highest accuracy (81.5%) and lowest AIC/BIC values, indicating superior performance. The log likelihood for BDML is (-1534.2), Bayesian Mixed Logit (BML) is -1541.1 and Mixed Logit (ML) is given as -1551.9 BDML model's best fit the data. The implications are that travel time and cost are significant factors in mode choice. The study recommended investments in comfortable and eco-friendly transportation and encourages bike usage through infrastructure development like good roads.