Inferring directional connectivity from point process data of multiple elements is desired in various scientific fields such as neuroscience, geography, economics, etc. Here, we propose an inference procedure for this goal based on the kinetic Ising model. The procedure is composed of two steps: (1) determination of the time-bin size for transforming the point-process data to discrete time binary data and (2) screening of relevant couplings from the estimated networks. For these, we develop simple methods based on information theory and computational statistics. Applications to data from artificial and \textit{in vitro} neuronal networks show that the proposed procedure performs fairly well when identifying relevant couplings, including the discrimination of their signs, with low computational cost. These results highlight the potential utility of the kinetic Ising model to analyze real interacting systems with event occurrences.