LOB modeling using Hawkes processes with a state-dependent factor

A point process model for order flows in limit books is proposed, which the conditional intensity product of a Hawkes component and state-dependent factor. In LOB context, state observations may include observed imbalance or spread. Full technical details computationally-efficient estimation such are provided, using either direct likelihood maximization EM-type estimation. Applications models bid ask market orders, upwards downwards price movements. Empirical results on multiple stocks traded Euronext Paris underline benefits formulations modeling, e.g. terms goodness-of-fit to financial data.