Modeling bi-directional effects between events and system variables

A computer-implemented method, comprising:

• learning, by a system operatively coupled to a memory and a processor, mutual dependencies jointly over event occurrence data and transition data, wherein the transition data comprises state variable transitions observed over a multivariate state variable set, wherein the learning comprises generating a state variable graphical event model that represents the mutual dependencies, the state variable graphical event model being a directed graph comprising:
• first nodes representing events,
• second nodes representing the state variable transitions, the second nodes being represented distinctly from the first nodes,
• first directed edges running from the first nodes to the second nodes and representing respective dependencies of certain state variable transitions on the events, and
• at least one second directed edge running from at least one of the second nodes to at least one of the first nodes and representing respective dependencies of the events on the state variable transitions,
• wherein the event occurrence data comprises time-stamped, asynchronous, irregularly spaced event data regarding multiple event types and the transition data is represented as the state variable transitions as opposed to event variables;
• receiving ongoing event occurrence data and ongoing transition data in real-time;
• generating, by the system, at least one of a predictive alert and a control action recommendation based on the received ongoing event occurrence data, the received ongoing transition data, and the state variable graphical event model;
• and
• presenting, by the system, the at least one of the predictive alert and the control action recommendation.

A computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to:

• learn mutual dependencies jointly over event occurrence data and transition data, wherein the transition data comprises state variable transitions observed over a multivariate state variable set, wherein the learning comprises generation of a state variable graphical event model that represents the mutual dependencies, the state variable graphical event model being a directed graph comprising:
• first nodes representing events,
• second nodes representing the state variable transitions, the second nodes being represented distinctly from the first nodes,
• first directed edges running from the first nodes to the second nodes and representing respective dependencies of certain state variable transitions on the events, and
• at least one second directed edge running from at least one of the second nodes to at least one of the first nodes and representing respective dependencies of the events on the state variable transitions,
• wherein the event occurrence data comprises time-stamped, asynchronous, irregularly spaced event data regarding multiple event types and the transition data is represented as the state variable transitions as opposed to event variables;
• receive ongoing event occurrence data and ongoing transition data in real-time;
• generate at least one of a predictive alert and a control action based on the received ongoing event occurrence data, the received ongoing transition data, and the state variable graphical event model;
• and
• present the at least one of the predictive alert and the control action.