Multidimensional user feature driven interface interaction design method and system

Abstract

The invention relates to the technical field of interface interaction design, in particular to a multi-dimensional user feature driven interface interaction design method and system. An interface interaction design system driven by multi-dimensional user characteristics comprises a mobile phone instability state judgment module, a track prediction module, an intention control determination module and an intention control amplification module. According to the method, the sliding track generated by the user interaction operation in the unstable state environment can be predicted to obtain the predicted track end point, and then the candidate controls are determined based on the predicted track end point; in combination with multi-dimensional user characteristics such as a historical use frequency ratio and functional logic correlation strength between the user and the candidate controls, an intention control most likely to interact next by the user is determined, and amplification operation setting is performed for the intention control, so that interaction between the user and the intention control is more convenient; and the interface interaction experience of the user in the unstable state environment is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of interface interaction design, in particular to a multi-dimensional user feature driven interface interaction design method and system. BACKGROUND

[0002] Current mobile terminal interface interaction design is still mainly optimized for static and stable environment, ignoring the user needs and behavior characteristics in dynamic scenarios. Especially on the subway, buses and other public transportation tools, due to the vibration, acceleration and deceleration and turning of the vehicle operation, users are difficult to maintain body balance, resulting in handheld devices unstable, in this unstable environment, the touch error rate is increased, and the user operation accuracy is decreased. Operation failure, task interruption and repeated operation frequently occur, and the user experience is poor. SUMMARY

[0003] The present application can predict the sliding trajectory generated by the user interaction operation in the unstable state environment by executing the mobile phone unstable interface interaction mode for the user interaction operation in the unstable state environment, obtain the predicted trajectory endpoint, and the predicted trajectory endpoint is the position where the user will interact with the mobile phone at the next monitoring time point. Then determine the candidate control based on the predicted trajectory endpoint, determine the intended control which the user is most likely to interact next by combining the multi-dimensional user features such as historical use frequency proportion and functional logical correlation strength between the user and the candidate control, and set the magnification operation for the intended control. Make the interaction between the user and the intended control more convenient, and improve the user interface interaction experience in the unstable state environment.

[0004] The present application provides a multi-dimensional user feature driven interface interaction design method, comprising: judging whether the mobile phone is in an unstable state, and if the mobile phone is in an unstable state, starting a mobile phone unstable interface interaction mode; In the phone unstable interface interaction mode, the user interaction position data of the current monitoring time point is obtained in response to the touch operation between the user and the phone interface, the user interaction position data is the X value and Y value corresponding to the position touched by the user on the phone interface, the user interaction position data of the current monitoring time point and the user interaction position data corresponding to the previous N-1 monitoring time points are combined to form a user interaction position time sequence set, the user interaction position time sequence set is sent to a trajectory prediction model for processing, and a predicted trajectory endpoint corresponding to the next monitoring time is output, a candidate control set is obtained based on the predicted trajectory endpoint, the candidate control set includes a plurality of candidate controls, the candidate control is a control within a preset distance from the predicted trajectory endpoint on the current phone interface, the intention score corresponding to all candidate controls in the candidate control set is calculated based on the distance-related value, the historical use frequency proportion and the functional logicality-related strength of the candidate control, and the candidate controls in the candidate control set are arranged in descending order of the corresponding intention score, the first N candidate controls are selected as intention controls, and an amplification operation is performed on the intention controls. The trajectory prediction model is established based on an LSTM model. Compared with the commonly used LSTM model, the trajectory prediction model adds reinforced updating of the forgetting gate weight matrix and the input gate weight matrix in the LSTM model.

[0005] As a preferred aspect, the reinforced updating of the forgetting gate weight matrix and the input gate weight matrix in the LSTM model in the trajectory prediction model specifically includes the following steps: The trajectory type data vector is determined based on the user interaction position time sequence set, and the trajectory type data vector includes the standard deviation of the speed, the standard deviation of the acceleration and the standard deviation of the direction change rate. The self-attention mechanism is performed on the forgetting gate weight matrix based on the trajectory type data vector, the corresponding value vector and key vector are constructed based on the forgetting gate weight matrix, and the corresponding query vector is constructed based on the trajectory type data vector. The self-attention mechanism is performed on the input gate weight matrix based on the trajectory type data vector, the corresponding value vector and key vector are constructed based on the input gate weight matrix, and the corresponding query vector is constructed based on the trajectory type data vector.

[0006] As a preferred aspect, the intention score corresponding to all candidate controls in the candidate control set is calculated based on the distance-related value, the historical use frequency proportion and the functional logicality-related strength of the candidate control, and specifically includes the following steps: The following operations are performed for each candidate control: the distance between the center of the candidate control and the end point of the predicted trajectory is calculated, denoted as a distance-related value, the ratio of the usage frequency of the candidate control to the usage frequency of all candidate controls in the candidate control set is calculated, denoted as a historical usage frequency proportion, the component corresponding to the last completed interaction operation is determined through the operation log, denoted as a target component, the target component and the candidate control form a function pair (target component, candidate control), the function pair is matched with the logical function library, if the matching is successful, the functional logical correlation strength is output as 1, if the matching fails, the functional logical correlation strength is output as 0, and the logical function library includes a plurality of logical function pairs; the distance-related value, the historical usage frequency proportion, and the functional logical correlation strength of the candidate control are combined to form an intention analysis vector, and the intention analysis vector is sent to the intention analysis network for processing to output the intention score corresponding to the candidate control.

[0007] As a preferred aspect, it also includes real-time training of the intention analysis network, and the specific steps are as follows: When the user performs an interaction operation on the mobile phone interface, the component that performs the interaction operation is recorded as a training component, the intention analysis vector corresponding to the training component at the last monitoring time point is obtained, denoted as a training intention analysis vector, the training intention analysis vector is sent to the current intention analysis network for training, during the training, the training intention analysis vector is used as the input of the intention analysis network, the predicted data amount output by the intention analysis network is obtained, the real-time loss value is calculated based on the difference between the predicted data amount and 1, and the intention analysis network is adjusted in parameters through the gradient descent method based on the real-time loss value, to realize real-time training of the intention analysis network.

[0008] As a preferred aspect, an amplification operation setting is performed for the intention control, specifically including the following steps: the amplification operation setting refers to amplifying the intention control to (1+A) times the original size, and A is the intention score corresponding to the intention control.

[0009] As a preferred aspect, the trajectory prediction model is trained, specifically including the following steps: A plurality of trajectory prediction training samples are obtained, the trajectory prediction training samples include N+1 user interaction position data, all trajectory prediction training samples are combined to form a trajectory prediction training set, the trajectory prediction model is trained through the trajectory prediction training set, during the training, the first N user interaction position data in the trajectory prediction training sample is used as the input of the trajectory prediction model, the last user interaction position data in the trajectory prediction training sample is used as the target output of the trajectory prediction model, the trajectory prediction loss value is calculated, it is judged whether the trajectory prediction loss value is located in the first preset range, if the trajectory prediction loss value is located in the first preset range, the trained trajectory prediction model is output; otherwise, the trajectory prediction model is continuously trained through the trajectory prediction training set.

[0010] As a preferred aspect, the intention analysis network is trained, specifically including the following steps: A plurality of intention analysis training samples are obtained, the intention analysis training samples include intention analysis vectors, the intention analysis training samples are labeled by intention scores, the labeled intention scores are all 1, all the labeled intention analysis training samples are combined to form an intention analysis training set, the intention analysis network is trained by the intention analysis training set, during the training, the intention analysis training samples are taken as the input of the intention analysis network, the intention scores labeled by the intention analysis training samples are taken as the target output of the intention analysis network, an intention analysis loss value is calculated, it is judged whether the intention analysis loss value is located in a second preset range, if the intention analysis loss value is located in the second preset range, the trained intention analysis network is output; otherwise, the intention analysis network is continuously trained by the intention analysis training set.

[0011] The application further provides a multi-dimensional user feature driven interface interaction design system, comprising: A mobile phone instability state judgment module is configured to judge whether the mobile phone is in an instability state, and start a mobile phone instability interface interaction mode if the mobile phone is in the instability state. A trajectory prediction module is configured to obtain user interaction position data at a current monitoring time point in response to a touch operation between the user and the mobile phone interface, the user interaction position data being X and Y values corresponding to positions touched by the user on the mobile phone interface, and to combine the user interaction position data at the current monitoring time point and user interaction position data corresponding to previous N-1 monitoring time points to form a user interaction position time sequence set, and to input the user interaction position time sequence set into a trajectory prediction model for processing to output a predicted trajectory endpoint corresponding to a next monitoring time. An intention control determination module is configured to obtain a candidate control set based on the predicted trajectory endpoint, the candidate control set including a plurality of candidate controls, the candidate controls being controls on the current mobile phone interface within a preset distance from the predicted trajectory endpoint, to calculate intention scores corresponding to all candidate controls in the candidate control set based on distance-related values, historical use frequency proportions and functional logic-related strengths of the candidate controls, and to arrange the candidate controls in the candidate control set in descending order of the corresponding intention scores, and select the first N candidate controls as intention controls. An intention control magnification module is configured to perform a magnification operation on the intention controls.

[0012] The application has the following advantages: The application can predict the sliding track generated by the user interaction operation in the unstable state environment, obtain the predicted track endpoint, and determine the candidate control based on the predicted track endpoint. The application combines the historical use frequency proportion and the function logic correlation strength between the user and the candidate control to determine the intended control that the user is most likely to interact with next, and sets the intended control for magnification operation, so that the interaction between the user and the intended control is more convenient, and the interface interaction experience of the user in the unstable state environment is improved. BRIEF DESCRIPTION OF DRAWINGS

[0013] Figure 1 The structure diagram of the multi-dimensional user feature driven interface interaction design system used in the embodiment of the application. DETAILED DESCRIPTION

[0014] In order to enable personnel in the art to better understand the technical solutions in the application, the technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application.

[0015] Embodiment 1, a multi-dimensional user feature driven interface interaction design method, comprising: The phone state data corresponding to the state of the phone is collected by the built-in gyroscope and accelerometer of the phone. The phone state data includes three-axis acceleration data and three-axis angular velocity. These data can represent the motion state of the phone in the user's hand, and further determine whether the phone is in an unstable state. The unstable state can be the unstable state of the user's center of gravity when the user is on the subway or standing on the bus. In this state, the interaction between the user and the phone interface will be affected by the unstable state, resulting in the user being unable to accurately perform the interaction operation. The phone state data is sent to the phone state analysis model for processing, and the phone state label is output. The phone state label includes a stable state label and an unstable state label. If the phone state label is a stable state label, the phone is not in an unstable state, and the next time stamp phone state data collection operation is continued. If the phone state label is an unstable state label, the phone is in an unstable state, and the phone unstable interface interaction mode is started. The phone state analysis model is established based on the BP neural network. In the phone unstable interface interaction mode, the user interaction position data of the current monitoring time point is obtained in response to the touch operation between the user and the phone interface. The user interaction position data here is the X value and Y value corresponding to the position touched by the user on the phone interface. The coordinate system corresponding to the X value and Y value here generally takes the left lower corner of the vertical interface of the phone as the origin, the horizontal right direction as the X axis direction, and the vertical upward direction as the Y axis direction. The user interaction position data of the current monitoring time point and the user interaction position data corresponding to the previous N-1 monitoring time points are combined to form a user interaction position time sequence set. If the user wants to interact with the phone, the user's thumb will slide with the phone in the case of phone instability. The sliding track can be used to predict the position where the user will interact with the phone next. The user interaction position time sequence set is sent to the trajectory prediction model for processing, and the predicted trajectory endpoint corresponding to the next monitoring time is output. The predicted trajectory endpoint is output in the form of X value and Y value. Based on the predicted trajectory endpoint, a candidate control set is obtained. The candidate control set includes a plurality of candidate controls. The candidate control is a control within a preset distance from the predicted trajectory endpoint on the current phone interface. The preset distance is set by the developer. The control is a component on the phone interface that can be interacted to achieve a function, such as copy, paste, and forward in the chat interface, add to shopping cart and view shopping cart in the purchase interface, etc. Based on the distance-related value, historical usage frequency proportion, and functional logicality-related strength of the candidate controls, the intention scores corresponding to all candidate controls in the candidate control set are calculated. The historical usage frequency proportion and the functional logicality-related strength are both based on the user's interaction behavior in the historical record, reflecting the multi-dimensional user features corresponding to the user's interaction operation with the phone interface. The intention score represents the probability that the user will perform an interaction on the control next. The candidate controls in the candidate control set are arranged in descending order of the corresponding intention scores. The top N candidate controls are selected as the intention controls. The intention controls are the controls that the user is most likely to perform an interaction operation on next. An amplification operation setting is performed on the intention controls. In order to make it more convenient for the user to perform an interaction operation on the intention controls, an amplification operation setting can be used to make the interaction between the user and the intention controls more convenient. The trajectory prediction model is established based on an LSTM model. Compared with the commonly used LSTM model, the trajectory prediction model adds reinforced updating of the forgetting gate weight matrix and the input gate weight matrix in the LSTM model. The LSTM model generally includes LSTM units as the basis, and each time of time series analysis is to send user interaction position data in a user interaction position time series set into the LSTM unit for processing, specifically including forgetting gate processing, input gate processing and output gate processing. The forgetting gate processing is to determine which information is discarded through the forgetting gate weight matrix, the input gate processing is to determine which information is retained, and the output gate processing is to determine which information is output. The specific forgetting gate processing, input gate processing and output gate processing operations refer to the existing LSTM model, and will not be described here. The trajectory prediction model in the present application reinforces the updating of the forgetting gate weight matrix and the input gate weight matrix in the commonly used LSTM model, and highlights the weight proportion of different trajectory prediction modes in trajectory prediction. The present application can predict the sliding trajectory generated by the user interaction operation in the unstable state environment by executing the mobile phone unstable interface interaction mode for the user interaction operation in the unstable state environment, obtain the predicted trajectory endpoint, and determine the candidate control based on the predicted trajectory endpoint. The predicted trajectory endpoint is the position where the user will interact with the mobile phone at the next monitoring time point. The multi-dimensional user features such as the historical use frequency proportion and the functional logical correlation strength between the user and the candidate control are combined to determine the intended control which the user is most likely to interact with next, and the intended control is set to be enlarged, so that the interaction between the user and the intended control is more convenient, and the interface interaction experience of the user in the unstable state environment is improved.

[0016] In the trajectory prediction model, the reinforced updating of the forgetting gate weight matrix and the input gate weight matrix in the LSTM model specifically includes the following steps: The trajectory type data vector is determined based on the user interaction position time sequence set, and the trajectory type data vector includes the standard deviation of the speed, the standard deviation of the acceleration and the standard deviation of the direction change rate. The standard deviation of the speed is determined based on the user interaction position time sequence set as follows: the distance between the user interaction position data corresponding to adjacent monitoring time points is divided by the time difference to obtain the tangential speed corresponding to each monitoring time point, and then the standard deviation of the speed is calculated based on the tangential speed corresponding to all the monitoring time points. The standard deviation of the acceleration is determined based on the user interaction position time sequence set as follows: the difference between the tangential speeds corresponding to adjacent monitoring time points is divided by the time difference to obtain the acceleration corresponding to each monitoring time point, and then the standard deviation of the acceleration is calculated based on the acceleration corresponding to all the monitoring time points. The standard deviation of the direction change rate is determined based on the user interaction position time sequence set as follows: the tangent direction corresponding to the monitoring time point is determined based on the vector between the user interaction position data corresponding to the monitoring time point and the user interaction position data corresponding to the next monitoring time point, the direction change rate is determined based on the tangent directions corresponding to adjacent monitoring time points divided by the corresponding time difference, and finally the standard deviation of all the direction change rates is calculated. The self-attention mechanism is performed on the forgetting gate weight matrix based on the trajectory type data vector, the corresponding value vector and key vector are constructed based on the forgetting gate weight matrix, and the corresponding query vector is constructed based on the trajectory type data vector. The trajectory type represented by the trajectory type data vector is used to guide which information is discarded. For example, if the standard deviation of the speed, the standard deviation of the acceleration and the standard deviation of the direction change rate in the trajectory type data vector are all low, it indicates that the sliding trajectory corresponding to the trajectory type data vector is a linear stable trajectory, and the probability of discarding the information of nonlinear trajectory prediction is increased. The self-attention mechanism is performed on the input gate weight matrix based on the trajectory type data vector, the corresponding value vector and key vector are constructed based on the input gate weight matrix, and the corresponding query vector is constructed based on the trajectory type data vector. The trajectory type represented by the trajectory type data vector is used to guide which information is retained. For example, if the standard deviation of the speed, the standard deviation of the acceleration and the standard deviation of the direction change rate in the trajectory type data vector are all low, it indicates that the sliding trajectory corresponding to the trajectory type data vector is a linear stable trajectory, and the probability of retaining the information of linear trajectory prediction is increased. The self-attention mechanism refers to the setting of the Transformer model, which generally performs matrix multiplication operations on the input vector and the value weight matrix, the key weight matrix and the query weight matrix to construct the corresponding value vector V, the key vector K and the query vector Q. The self-attention mechanism is realized through the following formula: H = softmax(QK T / D 0.5) V, wherein H is a vector output by the self-attention mechanism, T is a matrix transposition operation, and D is the dimension size of the key vector K; the value weight matrix, the key weight matrix, and the query weight matrix in the trajectory prediction model of the present application are trained end to end with the trajectory prediction model; The intention score corresponding to each candidate control in the candidate control set is calculated based on the distance-related value, the historical use frequency proportion, and the function logicality-related strength of the candidate control, and specifically includes the following steps: The following operations are performed for each candidate control: the distance between the center of the candidate control and the predicted trajectory endpoint is calculated, denoted as the distance-related value; the ratio of the use frequency of the candidate control to the use frequency of all candidate controls in the candidate control set is calculated, denoted as the historical use frequency proportion. It should be noted that the use frequency of each control is counted and recorded, and the component corresponding to the last completed interaction operation is determined through the operation log, denoted as the target component. The function pair (target component, candidate control) of the target component and the candidate control set is matched with the logical function library. If the matching is successful, the function logicality-related strength is output as 1, and if the matching fails, the function logicality-related strength is output as 0. The logical function library includes a plurality of logical function pairs (component, component), and the logical function library can be set by an operator or constructed through the operation log of the mobile phone. The distance-related value, the historical use frequency proportion, and the function logicality-related strength of the candidate control are combined to form an intention analysis vector, and the intention analysis vector is input into the intention analysis network for processing to output the intention score corresponding to the candidate control. The intention analysis network is based on a BP neural network; In order to improve the accuracy of the intention score analysis, real-time training of the intention analysis network is further included, and the specific steps are as follows: When the user performs an interaction operation on the mobile phone interface, the component that performs the interaction operation is recorded as a training component. The intention analysis vector corresponding to the training component at the last monitoring time point is obtained, denoted as a training intention analysis vector. The training intention analysis vector is input into the current intention analysis network for training. During the training, the training intention analysis vector is used as the input of the intention analysis network, and the predicted data amount output by the intention analysis network is obtained. The real-time loss value is calculated based on the difference between the predicted data amount and 1. The intention analysis network is adjusted in parameters through the gradient descent method based on the real-time loss value, realizing real-time training of the intention analysis network. It should be noted that the actual interaction component of the user is used as feedback to perform real-time training of the intention analysis network, which can make the analysis of the intention score more in line with the actual needs of the user, and further improve the interface interaction experience of the user in the unstable state environment; The following steps are included for setting the magnification operation of the intention control: the intention control is magnified to (1+A) times of the original size, and A is the intention score corresponding to the intention control. The mobile phone state analysis model is trained, specifically including the following steps: A plurality of mobile phone state analysis training samples are obtained, the mobile phone state analysis training samples including mobile phone state analysis vectors, the mobile phone state analysis vectors being actually collected by a developer on a mobile phone, the mobile phone state analysis training samples being labeled by mobile phone state labels, all the labeled mobile phone state analysis training samples being combined to form a mobile phone state analysis training set, the mobile phone state analysis model being trained by the mobile phone state analysis training set, during the training, the mobile phone state analysis training samples being used as input of the mobile phone state analysis model, the mobile phone state labels labeled on the mobile phone state analysis training samples being used as target output of the mobile phone state analysis model, whether the accuracy of the mobile phone state analysis model meets the expectation being determined, if the accuracy of the mobile phone state analysis model meets the expectation, the trained mobile phone state analysis model being output; otherwise, the mobile phone state analysis model being continuously trained by the mobile phone state analysis training set.

[0017] The trajectory prediction model is trained, specifically including the following steps: A plurality of trajectory prediction training samples are obtained, the trajectory prediction training samples including N+1 pieces of user interaction position data, the N+1 pieces of user interaction position data being actually collected in a simulated unstable environment, all the trajectory prediction training samples being combined to form a trajectory prediction training set, the trajectory prediction model being trained by the trajectory prediction training set, during the training, the first N pieces of user interaction position data in the trajectory prediction training samples being used as input of the trajectory prediction model, the last piece of user interaction position data in the trajectory prediction training samples being used as target output of the trajectory prediction model, a trajectory prediction loss value being calculated, whether the trajectory prediction loss value is within a first preset range being determined, the first preset range being set by a developer, if the trajectory prediction loss value is within the first preset range, the trained trajectory prediction model being output; otherwise, the trajectory prediction model being continuously trained by the trajectory prediction training set.

[0018] The intention analysis network is trained, specifically including the following steps: A plurality of intention analysis training samples are obtained, and the intention analysis training samples include intention analysis vectors, wherein the intention analysis vectors are constructed by the developer under a simulated unstable environment. After an interaction operation is performed, a component performing the interaction operation is recorded as a marked component. An intention analysis vector corresponding to the marked component at a previous monitoring time point is obtained and recorded as an intention analysis training sample. The intention analysis training sample is labeled by an intention score, wherein the intention score is 1. All labeled intention analysis training samples are combined to form an intention analysis training set. The intention analysis network is trained by using the intention analysis training set. During the training, the intention analysis training sample is used as the input of the intention analysis network, the intention score labeled by the intention analysis training sample is used as the target output of the intention analysis network, an intention analysis loss value is calculated, and it is determined whether the intention analysis loss value is within a second preset range. The second preset range is set by the developer. If the intention analysis loss value is within the second preset range, the trained intention analysis network is output. Otherwise, the intention analysis network is continuously trained by using the intention analysis training set.

[0019] In an embodiment 2, a multi-dimensional user feature driven interface interaction design system, as shown in Figure 1 , comprises: A mobile phone unstable state judgment module is configured to collect mobile phone state data corresponding to a state of the mobile phone by using a built-in gyroscope and an accelerometer of the mobile phone. The mobile phone state data includes three-axis acceleration data and three-axis angular velocity. These data can represent the motion state of the mobile phone in the user's hand and further determine whether the mobile phone is in an unstable state. The unstable state can be a state in which the user's center of gravity is unstable when the user is on a subway or standing on a bus. In this state, the interaction between the user and the mobile phone interface will be affected by the unstable state, resulting in the user being unable to accurately perform an interaction operation. The mobile phone state data is sent to a mobile phone state analysis model for processing, and a mobile phone state label is output. The mobile phone state label includes a stable state label and an unstable state label. If the mobile phone state label is the stable state label, the mobile phone is not in the unstable state, and the mobile phone state data collection operation of the next time stamp is continued. If the mobile phone state label is the unstable state label, the mobile phone is in the unstable state, and a mobile phone unstable interface interaction mode is started. The trajectory prediction module is configured to obtain user interaction position data at a current monitoring time point in response to a touch operation between the user and the mobile phone interface. The user interaction position data is X and Y values corresponding to a position touched by the user on the mobile phone interface. The coordinate system corresponding to the X and Y values has a left lower corner of the vertical interface of the mobile phone as an origin, the horizontal right direction as an X axis direction, and the vertical upward direction as a Y axis direction. The user interaction position data at the current monitoring time point and user interaction position data corresponding to N-1 previous monitoring time points are combined to form a user interaction position time sequence set. If the user wants to interact with the mobile phone, the user's thumb will slide with the mobile phone in the case of instability of the mobile phone. The sliding trajectory can be used to predict a position at which the user will interact with the mobile phone next. The user interaction position time sequence set is sent to a trajectory prediction model for processing, and a predicted trajectory endpoint corresponding to a next monitoring time is output. The predicted trajectory endpoint is output in the form of X and Y values. The intention control determination module is configured to obtain a candidate control set from the predicted trajectory endpoint. The candidate control set includes a plurality of candidate controls. The candidate controls are controls on the current mobile phone interface within a preset distance from the predicted trajectory endpoint. The preset distance is set by a developer. The controls are components on the mobile phone interface that can be interacted to realize functions, such as copy, paste, and forward in a chat interface, and adding to a shopping cart and viewing the shopping cart in a purchase interface. The intention scores corresponding to all candidate controls in the candidate control set are calculated based on distance-related values of the candidate controls, historical use frequency proportions, and functional logicality-related strengths. The historical use frequency proportions and the functional logicality-related strengths are both constructed based on interaction behaviors performed by the user in historical records, and reflect multi-dimensional user features corresponding to the interaction operation performed by the user on the mobile phone interface. The intention scores represent probabilities that the user will perform an interaction on the controls next. The candidate controls in the candidate control set are arranged in descending order of the corresponding intention scores. The first N candidate controls are selected as intention controls. The intention controls are controls on which the user is most likely to perform an interaction next. The intention control magnification module is configured to perform a magnification operation setting on the intention controls. In order to make it more convenient for the user to perform an interaction on the intention controls, the magnification operation setting can be used to make the interaction between the user and the intention controls more convenient.

[0020] It should be understood that those skilled in the art can make improvements or changes according to the above description, and all these improvements and changes shall fall within the protection scope of the appended claims of the present application. Parts not described in detail in the specification belong to the prior art known to those skilled in the art.

Claims

1. A multi-dimensional user feature driven interface interaction design method, characterized in that: include: Determine whether the phone is in an unstable state. If the phone is in an unstable state, activate the phone instability interface interaction mode; In the unstable interface interaction mode of the mobile phone, in response to the touch operation between the user and the mobile phone interface, the user interaction position data of the current monitoring time point is obtained. The user interaction position data is the X value and Y value corresponding to the position touched by the user on the mobile phone interface. The user interaction position data of the current monitoring time point and the user interaction position data corresponding to the previous N-1 monitoring time points are combined into a user interaction position time series set. The user interaction position time series set is sent to the trajectory prediction model for processing, and the predicted trajectory end point corresponding to the next monitoring time is output. Based on the predicted trajectory end point, a candidate control set is obtained. The candidate control set includes several candidate controls. The candidate control is a control on the current mobile phone interface that is within a preset distance from the predicted trajectory end point. The intention scores corresponding to all candidate controls in the candidate control set are calculated based on the distance correlation value, historical usage frequency ratio and functional logic correlation strength of the candidate control. The candidate control set is arranged from large to small according to the corresponding intention scores, the top N candidate controls are selected as intention controls, and the zoom operation setting is performed on the intention controls; The trajectory prediction model is built based on the LSTM model. Compared with the commonly used LSTM model, the trajectory prediction model adds enhanced updates to the forget gate weight matrix and input gate weight matrix in the LSTM model.

2. A multi-dimensional user feature driven interface interaction design method according to claim 1, characterized in that: In the trajectory prediction model, the enhanced update of the forget gate weight matrix and the input gate weight matrix in the LSTM model includes the following steps: Determine a trajectory type data vector based on the user interaction position time series set, where the trajectory type data vector includes a standard deviation of velocity, a standard deviation of acceleration, and a standard deviation of the direction change rate; Based on the trajectory type data vector, a self-attention mechanism is performed on the forget gate weight matrix, and the corresponding value vector and key vector are constructed based on the forget gate weight matrix, and the corresponding query vector is constructed based on the trajectory type data vector; Based on the trajectory type data vector, a self-attention mechanism is performed on the input gate weight matrix. The corresponding value vector and key vector are constructed based on the input gate weight matrix, and the corresponding query vector is constructed based on the trajectory type data vector.

3. The multi-dimensional user feature driven interface interaction design method according to claim 2, characterized in that: The intention scores of all candidate controls in the candidate control set are calculated based on their distance correlation values, historical usage frequency ratios, and functional logic correlation strengths. The specific steps are as follows: For each candidate control, the following operations are performed: the distance between the center of the candidate control and the end point of the predicted trajectory is calculated, which is recorded as the distance correlation value; the ratio of the number of times the candidate control is used to the number of times all candidate controls in the candidate control set are used is calculated, which is recorded as the historical usage frequency ratio; the component corresponding to the last completed interactive operation is determined through the operation log, which is recorded as the target component; the target component and the candidate control are combined into a function pair (target component, candidate control); the function pair is matched with the logical function library; if the match is successful, the output function logic correlation strength is 1; if the match fails, the output function logic correlation strength is 0; the logical function library includes several logical function pairs; the distance correlation value, historical usage frequency ratio and function logic correlation strength of the candidate control are combined into an intention analysis vector, which is then sent to the intention analysis network for processing, and the intention score corresponding to the candidate control is output.

4. The multi-dimensional user feature driven interface interaction design method according to claim 3, characterized in that: It also includes real-time training for the intent analysis network. The specific steps are as follows: When the user performs an interactive operation with the mobile phone interface, the component that performs the interactive operation is recorded as a training component, and the intention analysis vector corresponding to the training component corresponding to the previous monitoring time point is obtained and recorded as the training intention analysis vector. The training intention analysis vector is sent to the current intention analysis network for training. During the training, the training intention analysis vector is used as the input of the intention analysis network, and the predicted data volume output by the intention analysis network is obtained. The real-time loss value is calculated as the difference between the predicted data volume and 1. Based on the real-time loss value, the parameters of the intention analysis network are adjusted by the gradient descent method to realize real-time training of the intention analysis network.

5. The multi-dimensional user feature driven interface interaction design method according to claim 4, characterized in that: Perform the magnification operation setting for the intention control, which specifically includes the following steps: the magnification operation setting refers to magnifying the intention control to (1+A) times the original size, where A is the intention score corresponding to the intention control.

6. The multi-dimensional user feature driven interface interaction design method according to claim 5, characterized in that: Training the trajectory prediction model includes the following steps: Obtain several trajectory prediction training samples, each of which includes N+1 user interaction location data. All trajectory prediction training samples are combined into a trajectory prediction training set. The trajectory prediction model is trained using the trajectory prediction training set. During training, the first N user interaction location data in the trajectory prediction training samples are used as input to the trajectory prediction model, and the last user interaction location data in the trajectory prediction training samples is used as the target output of the trajectory prediction model. A trajectory prediction loss value is calculated to determine whether the trajectory prediction loss value is within a first preset range. If the trajectory prediction loss value is within the first preset range, the trained trajectory prediction model is output; otherwise, the trajectory prediction model is continued to be trained using the trajectory prediction training set.

7. The multi-dimensional user feature driven interface interaction design method according to claim 6, characterized in that: Training the intention analysis network includes the following steps: Obtain several intention analysis training samples, which include intention analysis vectors. Label the intention analysis training samples through intention scores, and the labeled intention scores are all 1. All labeled intention analysis training samples form an intention analysis training set, and train the intention analysis network through the intention analysis training set. During the training, the intention analysis training samples are used as the input of the intention analysis network, and the intention scores labeled by the intention analysis training samples are used as the target output of the intention analysis network. The intention analysis loss value is calculated to determine whether the intention analysis loss value is within a second preset range. If the intention analysis loss value is within the second preset range, output the trained intention analysis network; otherwise, continue to train the intention analysis network through the intention analysis training set.

8. A multi-dimensional user feature driven interface interaction design system, characterized in that: The system applies a multi-dimensional user feature-driven interface interaction design method according to any one of claims 1 to 7, comprising: The mobile phone instability state judgment module is used to judge whether the mobile phone is in an unstable state. If the mobile phone is in an unstable state, the mobile phone instability interface interaction mode is activated; The trajectory prediction module is used to respond to touch operations between the user and the mobile phone interface to obtain user interaction location data at the current monitoring time point. The user interaction location data is the X and Y values ​​corresponding to the location where the user touches the mobile phone interface. The user interaction location data at the current monitoring time point and the user interaction location data corresponding to the previous N-1 monitoring time points are combined into a user interaction location time series set. The user interaction location time series set is fed into the trajectory prediction model for processing and outputs the predicted trajectory endpoint corresponding to the next monitoring time. The intention control determination module is used to obtain a candidate control set based on the predicted trajectory endpoint. The candidate control set includes several candidate controls. The candidate control is a control within a preset distance from the predicted trajectory endpoint on the current mobile phone interface. The intention scores corresponding to all candidate controls in the candidate control set are calculated based on the distance correlation value, historical usage frequency ratio, and functional logic correlation strength of the candidate control. The candidate controls in the candidate control set are sorted from large to small according to the corresponding intention scores, and the top N candidate controls are selected and recorded as the intention controls; The intent control magnification module is used to perform magnification operation settings for intent controls.

Images