<p>Bike-sharing systems (BSSs) can help encourage a shift toward cycling. A key factor in implementing station-based BSSs is the location of the stations. Therefore, the stations’ locations should be optimized. Most of the previous studies considered bike-sharing demand static in location optimization. Further, all the presented studies considered a single-objective function. Equity has also not been considered in station location optimization at a city scale. To address these limitations, this study develops an accurate demand prediction model, which dynamically predicts the demand in the station location optimization process. Further, a multi-objective framework is developed to simultaneously maximize yearly bike-sharing use, maximize equity, and minimize implementation costs. A hybrid machine learning-optimization (LightGBM-RVEA) model is developed that optimizes station locations considering dynamic demand and many-objective functions. The model offers ten non-dominated optimal solutions by adding 7 to 223 stations to the network. In optimal solutions, increasing the number of added stations increases the number of trips on the network, but not with a similar improvement rate. Furthermore, all optimal solutions improve the city’s equity index. In the optimal solutions with more added stations, the improvement in the equity index is greater because we need more added stations to ameliorate the unequal distribution of BSS stations, and it may be impossible to address this issue with a few new stations. Moreover, the improvement of different solutions in terms of increasing the number of trips per added station and the optimal regions and their characteristics are discussed.</p>