This code can be used to reproduce the experiments of the ZLI-Team for the AI for TSP Competition. (see also: https://github.com/paulorocosta/ai-for-tsp-competition and https://www.tspcompetition.com) Specifically, this code tackles Track 1 (surrogate-based optimization) from that competition. In that track, one 55-node instance of the time-dependent orienteering problem with stochastic weights and time windows (TD-OPSWTW) has to be solved by using surrogate-based optimization algorithms. The solution approach combines (1) dimensionality reduction, (2) caching of solutions, (3) modeling the feasiblity of solutions via classification, (4) self-adaptive evolutionary optimization and (5) a surrogate model of the fitness function with ensemble-based Gaussian Process Regression.