Abstract: Current infectious disease surveillance studies often overlook the multi-layered nature of real-world contacts and disease characteristics. This paper presents a data-driven, multi-layer network approach that models disease transitions, including pre-symptomatic and asymptomatic states, using real infection data to estimate model parameters. We evaluate surveillance strategies based on early detection, peak delay, and peak ratio, and propose a collective influence strategy that considers contact weights and epidemiological features. Findings show that while traditional strategies can detect outbreaks about 3 days earlier, strategies accounting for network structure can advance this to 7 days, and the collective influence strategy can predict outbreaks up to two weeks in advance, demonstrating robustness to the size of the surveillance subset.