MTUSI: Filtering algorithm to estimate and predict changes in Poisson flow intensities

2022: Creating a filtering algorithm

Scientists from MTUSI December 22, 2022 announced the creation of algorithm filtering, which allows you to estimate and predict changes in the intensities of Poisson flows, as well as be included as an element in various control schemes, settings, identifications objects that meet the Poisson model.

A large range of events occurring in mass service systems can be described using the Poisson flow model. These include, for example, visits by customers to the store, acts of shopping, registration of subscribers in the cells of mobile communication systems, flows of requests for incoming and outgoing calls, requests for access to databases.

The intensity of these flows, as a rule, cannot be directly measured and does not remain constant, scientists say. Examples of situations that lead to accidental impacts on flow intensities are various kinds of accidents, unexpected actions on the part of competitors, unforeseen changes in conditions.

For the model of the Gaussian generating process in the Poisson flow intensity change equations, the solution to the filtration problem has already been obtained, but for the described cases with a high level of exposure to individual, rarely occurring random events, the known model is not suitable.

{{quote 'The resulting solution to the filtering problem has some benefit. Based on it, it is possible to determine the moments of occurrence and the scale of the impact of random events, as well as predict subsequent results. The synthesized algorithm can be used not only to estimate and predict changes in the intensities of Poisson flows, but also to be included as an element in various control schemes, settings, identification of objects that meet the Poisson model, - said Oleg Shorin, Doctor of Technical Sciences, Professor, Professor of the Department "Radio Engineering Systems" MTUSI. }}

The scientist noted that the analysis of the work of the synthesized algorithm was carried out using modeling, for which a program was written that forms a Poisson flow of events with changing intensity.

As a result, an algorithm for filtering the varying intensity of the observed Poisson event stream is obtained for a motion model including a generating process also in the form of a Poisson jump stream with a subsequent effect. It includes a detection unit and a unit for assessing the parameters of the sequence of jumps of the generating process, - said Oleg Shorin, Doctor of Technical Sciences, Professor, Professor of the Department of Radio Engineering Systems at MTUSI.

The filtering algorithm was tested, which confirmed the high quality of the filtering algorithm.