Highlights Introduces application Random Forest model for short-term demand prediction. GPS-based travel survey data is used to model activity travel of population segments. Correlations exist between socio-demographic attributes and travelers’ daily activity. Construct a series of ensemble learning based models to predict daily agenda.
Abstract The aim of this paper is to predict travel behavior for a set of model individuals, who represent cohorts with homogeneous time-use activity patterns. The paper presents a new modeling framework that is able to simulate temporal information associated with the traveler’s daily activity schedule, for use in activity-based travel demand modeling. We employed a precise and efficient machine learning technique known as Random-Forest for temporal attribute recognition. In order to capture the uncertainty of start time and activity duration, initially we derived unique clusters of homogeneous daily activity patterns from the activity data. The Random-Forest model is formulated based on the socio-demographic characteristics of travelers and temporal features of their activities. Start time and activity duration for every activity type were allocated to a set of bins. In this study, eight different bin structures, varying in the time interval, are designed as response variables. Using a heuristic decision rule-based algorithm, the predicted activities were inserted into the traveler’s skeleton schedule. An algorithm was then applied to schedule travelers’ activities based on activity importance level and guide information gained from representative patterns for homogeneous population cohorts. The model was tested using time-diary data drawn from the Space-Time Activity Research (STAR) survey for Halifax, Nova Scotia. Results show that the proposed model is able to assemble the traveler’s schedule with an average 81.62% accuracy in the 24-hour period. The insights gained from this study include important temporal information on activities crucial for the scheduling stage of an activity-based model. Finally, the results of this paper are expected to be implemented within the activity-based travel demand model, Scheduler for Activities, Locations, and Travel (SALT).
Ensemble learning activity scheduler for activity based travel demand models
2021-01-06
Article (Journal)
Electronic Resource
English
Activity-based travel demand modeling
TIBKAT | 1996
|Activity-Based Travel Demand Models to Evaluate Transport Policies
Springer Verlag | 2019
|Modeling Week Activity Schedules for Travel Demand Models
Transportation Research Record | 2017
|Activity Based Travel Demand Model Systems
Springer Verlag | 1998
|Activity-Based Modeling of Travel Demand
Springer Verlag | 1999
|