Accurate temporal event predictive modeling

Abstract

Embodiments for accurate temporal event predictive modeling by a processor. An average reverse event delay may be determined from one or more event delays in a time-series window. A time-series event may be predicted by applying the average reverse event delay in conjunction with one or more weighted factors in a predictive model.

1 . A method for accurate temporal event predictive modeling by a processor, comprising:
 identifying a plurality of time-series windows;   determining, for each time-series window, a reverse event delay, wherein the reverse event delay is an amount of time between the end of the time-series window and an occurrence of a corresponding hypoglycemic event;   determining, for each time-series window, an event delay, wherein the event delay is an amount of time between the start of the time-series window and the occurrence of the corresponding hypoglycemic event;   computing an average reverse event delay and an average event delay for the plurality of time-series windows;   training a predictive model for predicting hypoglycemic events using at least the average reverse event delay and the average event delay as weighted factors in the predictive model.   
     
     
         2 . The method of  claim 1 , further comprising grouping a plurality of patients into two or more patient groups,
 wherein a different predictive model is trained for each patient group using time-series event data for the respective patient groups.   
     
     
         3 . The method of  claim 2 , the method further comprising:
 determining a patient group that a particular patient is in;   identifying a particular predictive model associated with the patient group; and   predicting whether a future hypoglycemic event will occur for the particular patient using the particular predictive model.   
     
     
         4 . The method of  claim 1 , wherein the predictive model includes one or more of the group consisting of:
 an artificial neural network, a random forests model, an ensembles of classifiers, a bootstrap aggregating model, and a boosting model.   
     
     
         5 . The method of  claim 1 , wherein determining the reverse event delay for each of the plurality of time-series windows comprises:
 identifying a first time-series window, wherein the first time-series window is a window that starts at an occurrence of a first bolus event and ends a predetermined amount of time later;   determining whether a first hypoglycemic event occurs within the first time-series window;   if the first hypoglycemic event occurs within the first time-series window, calculating a first reverse event delay of the plurality of reverse event delays as an amount of time between an occurrence of the first hypoglycemic event and the end of the first time-series window; and   if the first time-series event does not occur within the first time-series window, setting the first reverse event delay as zero.   
     
     
         6 . The method of  claim 5 , wherein determining whether the first hypoglycemic events occurs within the first time-series window comprises:
 determining whether a blood glucose level of a patient drops below 70 mg/dl during the first time-series window;   if the blood glucose level of the patient drops below 70 mg/dl during the first time-series window, determining that the first time-series event occurred; and   if the blood glucose level of the patient does not drop below 70 mg/dl during the first time-series window, determining that the first time-series event did not occur.   
     
     
         7 . The method of  claim 1 , wherein training the predictive model comprises:
 deriving a plurality of features for the plurality of time-series windows, wherein the plurality of features include the average reverse event delay, the average event delay, and one or more other features;   clustering the plurality of features into two or more clusters using an unsupervised operation; and   training a machine learning model for each of the two or more clusters.   
     
     
         8 . A system comprising:
 a processor configured to perform a method comprising:   identifying a plurality of time-series windows;   determining, for each time-series window, a reverse event delay, wherein the reverse event delay is an amount of time between the end of the time-series window and an occurrence of a corresponding hypoglycemic event;   determining, for each time-series window, an event delay, wherein the event delay is an amount of time between the start of the time-series window and the occurrence of the corresponding hypoglycemic event;   computing an average reverse event delay and an average event delay for the plurality of time-series windows;   training a predictive model for predicting hypoglycemic events using at least the average reverse event delay and the average event delay as weighted factors in the predictive model.   
     
     
         9 . The system of  claim 8 , further comprising grouping a plurality of patients into two or more patient groups,
 wherein a different predictive model is trained for each patient group using time-series event data for the respective patient groups.   
     
     
         10 . The system of  claim 9 , wherein the method further comprises:
 determining a patient group that a particular patient is in;   identifying a particular predictive model associated with the patient group; and   predicting whether a future hypoglycemic event will occur for the particular patient using the particular predictive model.   
     
     
         11 . The system of  claim 8 , wherein the predictive model includes one or more of the group consisting of:
 an artificial neural network, a random forests model, an ensembles of classifiers, a bootstrap aggregating model, and a boosting model.   
     
     
         12 . The system of  claim 8 , wherein determining the reverse event delay for each of the plurality of time-series windows comprises:
 identifying a first time-series window, wherein the first time-series window is a window that starts at an occurrence of a first bolus event and ends a predetermined amount of time later;   determining whether a first hypoglycemic event occurs within the first time-series window;   if the first hypoglycemic event occurs within the first time-series window, calculating a first reverse event delay of the plurality of reverse event delays as an amount of time between an occurrence of the first hypoglycemic event and the end of the first time-series window; and   if the first time-series event does not occur within the first time-series window, setting the first reverse event delay as zero.   
     
     
         13 . The system of  claim 12 , wherein determining whether the first hypoglycemic events occurs within the first time-series window comprises:
 determining whether a blood glucose level of a patient drops below 70 mg/dl during the first time-series window;   if the blood glucose level of the patient drops below 70 mg/dl during the first time-series window, determining that the first time-series event occurred; and   if the blood glucose level of the patient does not drop below 70 mg/dl during the first time-series window, determining that the first time-series event did not occur.   
     
     
         14 . The system of  claim 8 , wherein training the predictive model comprises:
 deriving a plurality of features for the plurality of time-series windows, wherein the plurality of features include the average reverse event delay, the average event delay, and one or more other features;   clustering the plurality of features into two or more clusters using an unsupervised operation; and   training a machine learning model for each of the two or more clusters.   
     
     
         15 . A computer program product, the 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 perform a method comprising:
 identifying a plurality of time-series windows;   determining, for each time-series window, a reverse event delay, wherein the reverse event delay is an amount of time between the end of the time-series window and an occurrence of a corresponding hypoglycemic event;   determining, for each time-series window, an event delay, wherein the event delay is an amount of time between the start of the time-series window and the occurrence of the corresponding hypoglycemic event;   computing an average reverse event delay and an average event delay for the plurality of time-series windows;   training a predictive model for predicting hypoglycemic events using at least the average reverse event delay and the average event delay as weighted factors in the predictive model.   
     
     
         16 . The computer program product of  claim 15 , further comprising grouping a plurality of patients into two or more patient groups,
 wherein a different predictive model is trained for each patient group using time-series event data for the respective patient groups.   
     
     
         17 . The computer program product of  claim 16 , wherein the method further comprises:
 determining a patient group that a particular patient is in;   identifying a particular predictive model associated with the patient group; and   predicting whether a future hypoglycemic event will occur for the particular patient using the particular predictive model.   
     
     
         18 . The computer program product of  claim 15 , wherein determining the reverse event delay for each of the plurality of time-series windows comprises:
 identifying a first time-series window, wherein the first time-series window is a window that starts at an occurrence of a first bolus event and ends a predetermined amount of time later;   determining whether a first hypoglycemic event occurs within the first time-series window;   if the first hypoglycemic event occurs within the first time-series window, calculating a first reverse event delay of the plurality of reverse event delays as an amount of time between an occurrence of the first hypoglycemic event and the end of the first time-series window; and   if the first time-series event does not occur within the first time-series window, setting the first reverse event delay as zero.   
     
     
         19 . The computer program product of  claim 18 , wherein determining whether the first hypoglycemic events occurs within the first time-series window comprises:
 determining whether a blood glucose level of a patient drops below 70 mg/dl during the first time-series window;   if the blood glucose level of the patient drops below 70 mg/dl during the first time-series window, determining that the first time-series event occurred; and   if the blood glucose level of the patient does not drop below 70 mg/dl during the first time-series window, determining that the first time-series event did not occur.   
     
     
         20 . The computer program product of  claim 15 , wherein training the predictive model comprises:
 deriving a plurality of features for the plurality of time-series windows, wherein the plurality of features include the average reverse event delay, the average event delay, and one or more other features;   clustering the plurality of features into two or more clusters using an unsupervised operation; and   training a machine learning model for each of the two or more clusters.