Event processing based on multiple time windows
Implementations of the present disclosure relate to methods, devices, and computer program products for event processing. In the method, first data associated with a first time window is obtained, the first data comprising a first object and a first group of events that are related to the first object. Second data associated with a second time window is obtained, the second data comprising a second object and a second group of events that are related to the second object, the second time window being different from the first time window. An event model describing an association relationship between an object and an event that is related to the object is determined based on the first and second data. With these implementations, multiple time windows are used in determining the event model, and thus the event model may have better performance in accuracy and immediacy aspects.
1 . A method for event processing, comprising:
obtaining first data associated with a first time window, the first data comprising first objects and first groups of events, wherein each of the first groups of events is related to a corresponding first object among the first objects;
obtaining second data associated with a second time window, the second data comprising second objects and second groups of events, wherein each of the second groups of events is related to a corresponding second object among the second objects, and a duration of the second time window is shorter than a duration of the first time window;
generating first training samples based on the first data and generating second training samples based on the second data; and
training an initial model on the first training samples and the second training samples to obtain an event model, wherein the event model is configured to predict whether a particular event is to be received from a user in response to an object being displayed to the user.
2 . The method of claim 1 , wherein obtaining the first data comprises:
identifying the first object in response to a determination that the first object is sent within the first time window; and
selecting the first group of events in response to a determination that the first group of events are received within the first time window.
3 . The method of claim 1 , wherein generating a first sample among the first training samples comprises:
determining a first label for indicating whether the first group of events comprises a conversion event after a start event; and
generating the first sample based on the first data and the first label.
4 . The method of claim 3 , wherein determining the first label comprises:
setting the first label to be negative in response to a determination that the first group of events comprises the start event; and
updating the first label to be positive in response to a determination that the first group of events comprises the conversion event after the start event.
5 . The method of claim 3 , wherein the event model comprises:
a first model between an object and a first prediction of whether a conversion event related to the object is to be received within the first time window; and
a second model between an object and a second prediction of whether a conversion event related to the object is to be received within the second time window.
6 . The method of claim 5 , wherein determining the event model by training the initial model with the first and second samples comprises:
determining the first model by training a first portion in the initial model with the first sample; and
determining the second model by training a second portion in the initial model with the second sample.
7 . The method of claim 6 , wherein the event model further comprises: a weight model describing an association relationship between a first weight for the first time window and a second weight for the second time window, and determining the event model further comprises: determining the weight model based on the first and second samples.
8 . The method of claim 7 , further comprising: in response to a determination that a start event is related to a target object, determining a predication of whether a conversion event related to the target object is to be received based on the event model.
9 . The method of claim 8 , wherein determining the predication comprises:
receiving a first prediction from the first model, a second prediction from the second model, and a weight from the weight model based on the target object and the start event, respectively; and
determining the prediction based on the first and second predictions and the weight.
10 . An electronic device, comprising a computer processor coupled to a computer-readable memory unit, the memory unit comprising instructions that when executed by the computer processor cause the computer processor to perform operations, the operations comprising:
obtaining first data associated with a first time window, the first data comprising first objects and first groups of events, wherein each of the first groups of events is related to a corresponding first object among the first objects;
obtaining second data associated with a second time window, the second data comprising second objects and second groups of events, wherein each of the second groups of events is related to a corresponding second object among the second objects, and a duration of the second time window is shorter than a duration of the first time window;
generating first training samples based on the first data and generating second training samples based on the second data; and
training an initial model on the first training samples and the second training samples to obtain an event model, wherein the event model is configured to predict whether a particular event is to be received from a user in response to an object being displayed to the user.
11 . The device of claim 10 , wherein obtaining the first data comprises:
identifying the first object in response to a determination that the first object is sent within the first time window; and
selecting the first group of events in response to a determination that the first group of events are received within the first time window.
12 . The device of claim 10 , wherein generating a first sample among the first training samples comprises:
determining a first label for indicating whether the first group of events comprises a conversion event after a start event; and
generating the first sample based on the first data and the first label.
13 . The device of claim 12 , wherein determining the first label comprises:
setting the first label to be negative in response to a determination that the first group of events comprises the start event; and
updating the first label to be positive in response to a determination that the first group of events comprises the conversion event after the start event.
14 . The device of claim 12 , wherein the event model comprises:
a first model between an object and a first prediction of whether a conversion event related to the object is to be received within the first time window; and
a second model between an object and a second prediction of whether a conversion event related to the object is to be received within the second time window.
15 . The device of claim 14 , wherein determining the event model by training the initial model with the first and second samples comprises:
determining the first model by training a first portion in the initial model with the first sample; and
determining the second model by training a second portion in the initial model with the second sample.
16 . The device of claim 15 , wherein the event model further comprises: a weight model describing an association relationship between a first weight for the first time window and a second weight for the second time window, and determining the event model further comprises: determining the weight model based on the first and second samples.
17 . The device of claim 16 , the operations further comprising:
in response to a determination that a start event is related to a target object, determining a predication of whether a conversion event related to the target object is to be received based on the event model, comprising:
receiving a first prediction from the first model, a second prediction from the second model, and a weight from the weight model based on the target object and the start event, respectively; and
determining the prediction based on the first and second predictions and the weight.
18 . A computer program product, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by an electronic device to cause the electronic device to perform operations, the operations comprise:
obtaining first data associated with a first time window, the first data comprising first objects and first groups of events, wherein each of the first groups of events is related to a corresponding first object among the first objects;
obtaining second data associated with a second time window, the second data comprising second objects and second groups of events, wherein each of the second groups of events is related to a corresponding second object among the second objects, and a duration of the second time window is shorter than a duration of the first time window;
generating first training samples based on the first data and generating second training samples based on the second data; and
training an initial model on the first training samples and the second training samples to obtain an event model, wherein the event model is configured to predict whether a particular event is to be received from a user in response to an object being displayed to the user.