Testing control method and apparatus for application, and electronic device and storage medium

An electronic device, comprising:

• one or more processors;
• a memory;
• and
• one or more application programs, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, the one or more application are programs configured to:
• acquire, in response to an automated testing request for a target application, semantic segmentation maps of a current access page and at least two consecutive historical access pages associated with the current access page, and a historical operation type and a historical operation location map that correspond to each of the at least two consecutive historical access pages;
• input the semantic segmentation maps, the historical operation type, and the historical operation location map into a pre-trained behavior prediction model, and acquire a target operation type and a target operation location probability map on the current access page that are predicted by the pre-trained behavior prediction model, the target operation location probability map characterizing a probability of performing a target operation on a page location;
• and
• control a testing of the current access page of the target application based on the target operation type and the target operation location probability map,
• wherein the pre-trained behavior prediction model comprises a 3D convolutional neural network layer, a long short-term memory (LSTM) neural network layer and an output layer that are cascaded, wherein
• the 3D convolutional neural network layer is configured to extract spatial information of at least two consecutive access pages and time sequence information among respective access pages;
• the long short-term memory (LSTM) neural network layer is configured to learn time sequence information of different scales in the spatial information;
• and
• the output layer is configured to output the predicted operation type and the predicted operation location probability map, and
• wherein the one or more application programs are further configured to:
• extract first spatial information of the current access page and each of the at least two consecutive historical access pages using the 3D convolutional neural network layer in the pre-trained behavior prediction model, wherein the first spatial information comprises a location relationship and a size relationship among respective semantic segmentation regions in each of the current access page and the at least two consecutive historical access pages;
• extract first time sequence information among the current access page and the at least two consecutive historical access pages using the 3D convolutional neural network layer in the pre-trained behavior prediction model, wherein the first time sequence information comprises a first association among the current access page and the at least two consecutive historical access pages that is determined based on an access page occurrence sequence;
• learn second time sequence information among the current access page and the at least two consecutive historical access pages that is contained in the first spatial information using the LSTM network layer in the pre-trained behavior prediction model, wherein the second time sequence information comprises second associations on different scales among the current access page and the at least two consecutive historical access pages;
• output the target operation type on the current access page using the first association among the current access page and the at least two consecutive historical access pages or the second associations on different scales among the current access page and the at least two consecutive historical access pages;
• and
• output the target operation location probability map on the current access page using the location relationship and the size relationship among the respective semantic segmentation regions in the current access page and in each of the at least two consecutive historical access pages.

A non-transitory computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps of:

• acquiring, in response to an automated testing request for a target application, semantic segmentation maps of a current access page and at least two consecutive historical access pages associated with the current access page, and a historical operation type and a historical operation location map that correspond to each of the at least two consecutive historical access pages;
• inputting the semantic segmentation maps, the historical operation type, and the historical operation location map into a pre-trained behavior prediction model, and acquiring a target operation type and a target operation location probability map on the current access page that are predicted by the pre-trained behavior prediction model, the target operation location probability map characterizing a probability of performing a target operation on a page location;
• and
• controlling a testing of the current access page of the target application based on the target operation type and the target operation location probability map,
• wherein the pre-trained behavior prediction model comprises a 3D convolutional neural network layer, a long short-term memory (LSTM) neural network layer and an output layer that are cascaded, wherein
• the 3D convolutional neural network layer is configured to extract spatial information of at least two consecutive access pages and time sequence information among respective access pages;
• the LSTM neural network layer is configured to learn time sequence information of different scales in the spatial information;
• and
• the output layer is configured to output the predicted operation type and the predicted operation location probability map, and
• wherein said inputting the semantic segmentation maps, the historical operation type, and the historical operation location map into the pre-trained behavior prediction model, and acquiring the target operation type and the target operation location probability map on the current access page that are predicted by the pre-trained behavior prediction model comprises:
• extracting first spatial information of the current access page and each of the at least two consecutive historical access pages using the 3D convolutional neural network layer in the pre-trained behavior prediction model, wherein the first spatial information comprises a location relationship and a size relationship among respective semantic segmentation regions in each of the current access page and the at least two consecutive historical access pages;
• extracting first time sequence information among the current access page and the at least two consecutive historical access pages using the 3D convolutional neural network layer in the pre-trained behavior prediction model, wherein the first time sequence information comprises a first association among the current access page and the at least two consecutive historical access pages that is determined based on an access page occurrence sequence;
• learning second time sequence information among the current access page and the at least two consecutive historical access pages that is contained in the first spatial information using the LSTM network layer in the pre-trained behavior prediction model, wherein the second time sequence information comprises second associations on different scales among the current access page and the at least two consecutive historical access pages;
• outputting the target operation type on the current access page using the first association among the current access page and the at least two consecutive historical access pages or the second associations on different scales among the current access page and the at least two consecutive historical access pages;
• and
• outputting the target operation location probability map on the current access page using the location relationship and the size relationship among the respective semantic segmentation regions in the current access page and in each of the at least two consecutive historical access pages.