In this study, computational fluid dynamics (CFD) analyses of the two-dimensional, time-dependent lid-driven cavity flows, for Reynolds numbers ranging from 100 to 10000, are performed by using an in-house developed CFD code. The unsteady behavior of the flow is triggered using a sinusoidal lid velocity profile. The flow structure is further investigated with the application of a reduced order modeling technique, Proper Orthogonal Decomposition (POD), and the structures present in the flow, are separated according to their frequency (energy) content. POD results show that when the stream function formation is used as a data ensemble, about 99% of the total energy content can be modeled by considering only the most energetic first four POD modes; whereas, this value remains at a range between 90 - 95% for the x-direction velocity data ensemble. What is more, an Artificial Neural Network (ANN) based approach is developed to predict mode amplitudes for flows with different Reynolds numbers. Once enough information is obtained with the help of CFD of few flow cases, the ANN integrated approach presented herein helps to predict what is happening in the flow for different flow cases without requiring further CFD simulations, which are not practical in real-time flow control applications.