Method for establishing applications-to-be preloaded prediction model based on preorder usage sequence of foreground application, storage medium, and terminal

1. A method for establishing an application prediction model, comprising:

collecting a first preorder usage sequence of a first application currently running as sample usage information, in response to an operation switching an application previously running in the foreground to the first application;

monitoring applications launched within a time-window from a starting point of collecting the sample usage information and having a preset length, and setting a preset number of applications launched first in the applications launched within the time-window as sample labels for the sample usage information; and

training, based on the sample usage information and the sample labels for the sample usage information, a predetermined machine learning model to obtain an application prediction model,

wherein setting the preset number of applications launched first in the applications launched within the time-window as the sample labels for the sample usage information comprises:

allocating probabilities to the preset number of applications launched first in the applications launched within the time-window according to intervals from the start time point to launching time points of the applications launched first; and

setting the preset number of applications and the probabilities of the preset number of applications as the sample labels for the sample usage information.

2. The method of claim 1 , further comprising:

acquiring status feature information of a terminal corresponding to a time point at which the sample usage information is collected,

wherein training the predetermined machine learning model based on the sample usage information and the sample labels for the sample usage information comprises:

training the predetermined machine learning model based on the status feature information, the sample usage information, and the sample labels for the sample usage information.

3. The method of claim 1 , wherein the preset number is N, and the method further comprises:

acquiring a second preorder usage sequence of a second application currently running of a terminal, in response to an event of application preloading prediction being triggered;

inputting the second preorder usage sequence to the application prediction model;

predicting N target applications to be launched according to an output result of the application prediction model; and

preloading the N target applications.

4. The method of claim 2 , wherein the predetermined machine learning model is a hybrid network model, the hybrid network model comprises a recurrent neural network (RNN) sub-model and a fully-connected network sub-model built based on a fully connected layer, and an output layer of the RNN sub-model and an output layer of the fully-connected network sub-model build the hybrid network model based on a full connection relationship; and

wherein training the predetermined machine learning model based on the status feature information, the sample usage information, and the sample labels for the sample usage information comprises:

training the hybrid network model, by inputting the sample usage information and the sample labels for the sample usage information to the RNN sub-model and inputting the status feature information to the fully-connected network sub-model.

5. The method of claim 2 , wherein the status feature information comprises at least one of:

time information, a date category, an on/off status of a mobile data network, a connected/disconnected status of a wireless hotspot, identity information of a connected wireless hotspot, a duration for which a current application stays in the background, a time point at which the current application was last switched to the background, a plugging/unplugging status of an earphone jack, a charging status, power information of a battery, a display duration of a screen, and motion status and location information of the terminal.

6. The method of claim 3 , wherein preloading the N target applications comprises:

preloading application interfaces of the N target applications based on a pre-established preloading active-window stack, the pre-established preloading active-window stack having boundary coordinates located beyond the coordinate range of a display screen.

7. The method of claim 3 , wherein preloading the N target applications comprises:

recording the N target applications in a list of applications to be preloaded; and

preloading each of the target applications in the list of applications to be preloaded sequentially at an interval of a preset time period for preloading until all the N target applications are completely loaded.

8. The method of claim 3 , further comprising:

determining whether a launching operation for a predetermined code path of the second application is monitored; and

suspending preloading of the N target applications, in response to determining that the launching operation of the predetermined code path is monitored.

9. The method of claim 3 , further comprising one of:

acquiring central processing unit (CPU) utilization of the terminal; and determining that the CPU utilization is smaller than a preset utilization threshold; and

acquiring remaining storage capacity of the terminal; and determining that the remaining storage capacity is greater than a preset storage threshold.

10. A non-transitory computer-readable storage medium storing a computer program which, when executed by a processor, causes the processor to carry out actions, comprising:

collecting a first preorder usage sequence of a first application currently running as sample usage information, in response to an operation switching an application previously running in the foreground to the first application;

monitoring applications launched within a time-window from a starting point of collecting the sample usage information and having a preset length and setting a preset number of applications launched first in the applications launched within a time-window as sample labels for the sample usage information; and

training a predetermined machine learning model based on the sample usage information and the sample labels for the sample usage information to obtain an application prediction model,

wherein setting the preset number of applications launched first in the applications launched within the time-window as the sample labels for the sample usage information is executed by the processor to carry out actions comprises:

allocating probabilities to the preset number of applications launched first in the applications launched within the time-window according to intervals from the start time point to launching time points of the applications launched first; and

setting the preset number of applications and the probabilities of the preset number of applications as the sample labels for the sample usage information.

11. The non-transitory computer-readable storage medium of claim 10 , wherein the computer program is further executed by the processor to carry out actions, comprising:

acquiring status feature information of a terminal corresponding to a time point at which the sample usage information is collected,

wherein the computer program executed by the processor to carry out the action of training the predetermined machine learning model based on the sample usage information and the sample labels for the sample usage information is executed by the processor to carry out actions, comprising:

training the predetermined machine learning model based on the status feature information, the sample usage information, and the sample labels for the sample usage information.

12. The non-transitory computer-readable storage medium of claim 10 , wherein the computer program is further executed by the processor to carry out actions, comprising:

acquiring a second preorder usage sequence of a second application currently running of a terminal, in response to an event of application preloading prediction being triggered;

inputting the second preorder usage sequence to the application prediction model;

predicting N target applications to be launched according to an output result of the application prediction model; and

preloading the N target applications.

13. A terminal, comprising:

at least one processor; and

a computer readable storage, coupled to the at least one processor and storing at least one computer executable instruction thereon which, when executed by the at least one processor, cause the at least one processor to carry out actions, comprising:

collecting a first preorder usage sequence of a first application currently running as sample usage information, in response to an operation switching an application previously running in the foreground to the first application;

monitoring applications launched within a time-window from a starting point of collecting the sample usage information and having a preset length and setting a preset number of applications launched first in the applications launched within a time-window as sample labels for the sample usage information; and

training a predetermined machine learning model based on the sample usage information and the sample labels for the sample usage information to obtain an application prediction model,

wherein setting the preset number of applications launched first in the applications launched within the time-window as the sample labels for the sample usage information is caused to carry out actions comprises:

allocating probabilities to the preset number of applications launched first in the applications launched within the time-window according to intervals from the start time point to launching time points of the applications launched first; and

setting the preset number of applications and the probabilities of the preset number of applications as the sample labels for the sample usage information.

14. The terminal of claim 13 , wherein the at least one processor is further caused to carry out actions, comprising:

acquiring a second preorder usage sequence of a second application currently running of a terminal, in response to an event of application preloading prediction being triggered;

inputting the second preorder usage sequence to the application prediction model;

predicting N target applications to be launched according to an output result of the application prediction model; and

preloading the N target applications.