This study aims to implement two distinct approaches for evaluating the seismic vulnerability of Skikda city. The first approach is the traditional European macroseismic Risk-UE LM1 method, which involves in-situ surveys of building characteristics. However, this approach is time-consuming, costly, and necessitates the involvement of highly qualified personnel. The second approach utilizes a data-mining technique called Association Rule Learning (ARL) to minimize the need for extensive building attribute data. The ARL method associates building attributes with the European Macroseismic Scale (EMS-98) vulnerability classes obtained from visual observation surveys. We then validated the obtained vulnerability proxies in the Skikda database. While minor variations exist in the probability of exceeding the specified damage level, a comparison of Risk-UE LM1 and ARL results confirms the general reliability of the seismic vulnerability assessment. The damage estimates are evaluated against deterministic and probabilistic scenarios, considering moderate to severe damage and other impacts such as human casualties, direct economic costs, and debris volumes.