Group trajectories according to queries, hold trajectories conform to each query for test, and use trajectories conform to all other queries for training.
To evaluate a specific prediction $y$ w.r.t. query $x$, using the maximum (or average?) similarity score between $y$ and every trajectory in test set (i.e. trajectories conform to $x$).
Group trajectories according to queries, hold trajectories conform to each query for test, and use trajectories conform to all other queries for training.
To evaluate a specific prediction $y$ w.r.t. query $x$, using the maximum (or average?) similarity score between $y$ and every trajectory in test set (i.e. trajectories conform to $x$).