چكيده به لاتين
Urban planning and it’s relation with transportation is closely tied to understanding human activity patterns within urban areas. Consequently, modeling and predicting individual travel behavior have been essential for years. Historically, addressing long-term transportation needs involved creating transportation infrastructure based on these patterns. However, over time, due to rising costs of new infrastructure and concerns about traffic congestion and environment pollution impact, transportation planning has shifted focus. It now encompasses managing costs, accessibility challenges, and travel demand management.As a result, various travel demand management strategies for travel behavior such as congestion pricing, telecommuting, flexible work hours, and intelligent transportation systems aim to alter individual patterns and effectively manage travel behavior. The transition from traditional four-step travel demand modeling to disaggregate travel modeling (at the individual level) has been a significant development in travel planning.In this study, we predict specific activity patterns based on individuals’ socio-economic characteristics. We utilize data from the city of Washington, collected in 2007-2008. Using Generative Adversarial Networks (GANs) and Auto-Encoders, we input socio-economic information such as age, gender, income, eployment status and the model predicts individuals’ activities throughout the day in 48 thirty minutes intervals, categorizing them into six types: home, work, education, shopping, recreation, and other.To validate our models, we employ metrics such as Fréchet Inception Distance (FID), chi-suare test, pattern analysis, t-test and comparisons of total activity counts. Based on validation, the Wasserstein Generative Adversarial Network (WGAN) emerges as the best model. It accurately predicts individuals’ activities during a 24 hours of a day.
Keywords: Travel modeling, activity-based modeling, individual activity patterns, Generative Adversarial Network (GAN), Auto-Encoder
