Assessing Disparities in Road Transport Services in Kenya: Geographic Information System and Machine Learning Approach

Regional disparities in road infrastructure access in Kenya hinder socioeconomic development and challenge progress toward the Vision 2030 goals. This study addresses spatial inequality in transport provision by developing a transport network need index (TNNI) based on socioeconomic indicators and comparing it with existing road network density, termed transport network provision (TNP). The difference between TNNI and TNP yields the index of disparity between needs and provision (IDNP), identifying counties that are over- or under-served relative to their development needs. Machine learning models are employed to predict road density based on socioeconomic features, with the best-performing model selected through accuracy metrics. Shapley additive explanations (SHAP) are used to interpret the model and determine the empirical importance of each input feature. These SHAP-based weights are compared with literature-based weights used in TNNI to assess disparities in factor prioritization. Results highlight critical spatial disparities, with counties such as Turkana, Wajir, and Mandera falling into the most under-served category (IDNP >0.8). SHAP analysis reveals that population density is the dominant driver of current road provision (56.8% weight), while poverty and unemployment, despite their policy relevance, are underrepresented in actual infrastructure allocation. This twofold analysis—identifying both regional service gaps through IDNP and mismatches in factor prioritization—provides critical insights for policymakers. By pinpointing under-served regions and highlighting which socioeconomic drivers are being overlooked, the study offers a robust, data-driven framework for guiding equitable infrastructure investments, advancing transport justice, and supporting diversity and inclusion in Kenya’s road transport planning.