The potential of resistive switching (RS) devices based on TiN/Ti/HfO2/W stacks to mimic synapses within a neuromorphic applications context is analyzed in depth. The fabrication and characterization process are explained and a physically-based modeling description is performed to understand the devices resistive switching operation and conductance modulation. The model employed considers truncated-cone shaped conductive filament (CF) geometries and parasitic ohmic resistances linked to the device conductive filaments in addition to device capacitances. The temporal evolution is analysed assuming a valence change memory operation, where the oxide surrounding the CF is considered as well as the CF thermal description. A complete series of RS cycles has been fitted with the model by means of the gradient descent algorithm to study the compliance current effects on the conductance modulation. To do so, experimental and modeled results are extensively compared.