[Day 152] SVR & STTCM - Two architectures for taxi demand prediction
Hello :) Today is Day 152! A quick summary of today (papers read): Optimization of spatial-temporal graph: A taxi demand forecasting model based on spatial-temporal tree Predicting Taxi-Out Time at Congested Airports with Optimization-Based Support Vector Regression Methods Updated graph of read papers from Obsidian: The first paper introduces Spatial-Temporal Tree Convolution Model (STTCM) Introduction Spatial-temporal dependencies in taxi demand reflect changes across different locations and times. Temporally, demand follows city rhythms, being higher during rush hours and lower late at night, with variations between weekdays and weekends. Spatially, demand varies by area type, such as high demand in business districts during rush hours and in nightlife zones late at night. These dependencies interact; for instance, high morning demand in business districts is linked to the time of day, just as late-night demand in hotspots depends on it being late night. Taxi demand is influenced b