| CPC G06Q 30/016 (2013.01) [G06F 18/2155 (2023.01); G06N 20/00 (2019.01); G06Q 10/02 (2013.01)] | 12 Claims |
|
1. A system comprising:
a processor comprising hardware, the processor is configured to be in communication with the memory, and the processor is configured to:
receive a set of service records from a data source, one or more service records corresponding to a plurality of weather-related service disturbance-revealing events of a specified region occurring during a time period, one or more service records being mislabeled or having no label relating to an associated weather-related service disturbance;
determine an observed actual event rate for weather-related service disturbance-revealing events of the time period;
determine a baseline mean periodically-aggregated rate of expected service related records under non-disturbance conditions of the time period;
determine a set of standardized periodically-aggregated scores as a function of the baseline mean periodically-aggregated rate and observed actual event rate during the time period;
apply a calibrated change-point analysis to identify a weather-related service disturbance time window based on changes of score values in the set of standardized periodically-aggregated scores detected during the time period, said identifying comprising applying a calibrated change-point analysis to separate a non-disturbance condition of the time period from a disturbance condition of the time period, wherein the applying a calibrated change-point analysis to identify a disturbance time window comprises:
generate an auto-restarting cumulative sum chart having a sequence of control scheme values for disturbance identification, said sequence of control scheme values being transformed from corresponding score values in the set of standardized periodically-aggregated scores and modified according to a cumulative sum reference parameter;
identify a time when a control scheme value exceeds a sensitivity threshold parameter value;
determine an end of episode condition that establishes a bound for an end time of said disturbance condition and determine an end time of said disturbance time window based on the identified time in said auto-restarting cumulative sum chart;
determine a beginning of episode condition that establishes a bound for a start time of said disturbance condition and determine a start time of said disturbance time window based on the identified time in said auto-restarting cumulative sum chart; and
automatically re-start a transforming of said sequence of control scheme values from corresponding values at each identified time a control scheme value exceeds a sensitivity threshold parameter value; and
generate a disturbance-related probability for the service records corresponding to the identified weather-related service disturbance time window;
re-assign a label, based on the generated probability, to a service ticket as being related to the identified weather-related service disturbance time window;
obtain, based on re-assigned labels to service tickets for identified weather-related service disturbance time windows associated with past storms, a time period for the past storms;
monitor a current set of weather conditions, and
run a machine learning prediction model to predict, based on the current weather condition, a weather-related service disturbance event potentially affecting a number of customers or resources within a time frame of interest, said prediction model trained using the re-assigned service ticket labels to correlate past identified disturbance time windows associated with past storms with infrastructure and weather related variables for use in minimizing weather-related service disturbances;
wherein responsive to said prediction, implement a preventative action to mitigate any predicted weather-related service disturbance event in the specified region.
|