Analysis device, analysis program, analysis method, estimation device, estimation program, and estimation method

A computer program product for determining a scale of a response of one or more individuals to a future event, the computer program product comprising a non-transitory computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to perform a method comprising:

• inputting data representing a history of events associated with an individual, said history of events arranged in a time series over a period of time;
• substituting each event in the inputted event time series with rectangular windows of predetermined time width;
• dividing the event series timewise according to transition points corresponding to beginning timestamps and end timestamps associated with each of the rectangular windows to obtain multiple adjacent time periods;
• generating a state vector corresponding to the state of the rectangular windows in each of the multiple adjacent time periods;
• calculating a response frequency for each time period, the response frequency being the number of responses generated during the time period;
• converting the state vectors to feature vectors;
• performing a Poisson regression using the state vectors corresponding to the time periods,
• optimizing a weight vector to maximize a likelihood with which response frequencies can be obtained in the time periods from a Poisson distribution on the basis of a scalar score obtained from the scalar product of the feature vectors and the weight vector,
• storing in the memory, the optimized weight vector for the individual;
• and
• determining a response characteristic of the individual to a future event series using the stored weight vector.

An analyzing method for determining a scale of a response of many individuals to a future event, said method comprising the computer-executable steps of:

• receiving, using a processor, data representing a history of events associated with an individual, said history of events arranged in a time series over a period of time;
• substituting, using the processor, each event in each of the received one or more event time series with rectangular windows of pre-determined time width;
• dividing, using the processor, the event series timewise according to transition points corresponding to beginning timestamps and end timestamps associated with each of the rectangular windows to obtain multiple adjacent time periods;
• generating, using the processor, a state vector corresponding to the state of the rectangular windows in each of the multiple adjacent time periods;
• calculating, using the processor, a response frequency for each time period, the response frequency being the number of responses generated during the time period;
• converting, using the processor, the state vectors to feature vectors;
• performing, using the processor, a Poisson regression using the state vectors and the response frequencies corresponding to the time periods;
• and
• optimizing, using the processor, a weight vector to maximize the likelihood with which response frequencies can be obtained in the time periods from a Poisson distribution on the basis of a scalar score obtained from the scalar product of the feature vectors and the weight vector;
• storing in the memory, the optimized weight vector for the individual;
• and
• determining, using the processor, a response characteristic of the individual to a future event series using the stored weight vector.