Method, device and equipment for constructing decision prediction model based on visual information

Abstract

The invention relates to a method, a device and equipment for constructing a decision prediction model based on visual information. Comprising the steps that an eye movement data set is constructed, and the eye movement data set comprises visual attributes reflected by different eye movement data in the decision making process; obtaining a plurality of candidate models, and determining a target model from the plurality of candidate models according to the accuracy rates and Kappa coefficients corresponding to the plurality of candidate models; and determining a target hyper-parameter corresponding to the target model by using a ten-fold cross validation method, and training the target model by using the target hyper-parameter and the eye movement data set to obtain a decision prediction model. On the basis, the eye movement data can be predicted through the target model, so that more eye movement data can be quickly analyzed, a corresponding decision is generated, and the decision generation efficiency is improved.

Description

Technical Field

[0001] The present disclosure relates to the technical field of information management systems, and in particular to a method, device and equipment for constructing a decision prediction model based on visual information. Background Art

[0002] In recent years, with the rapid development of Internet technology, the user behaviors generated by users when using the Internet have become richer, and rich user behaviors will generate rich information.

[0003] In related technologies, decision makers can provide decisions based on information corresponding to user behavior, but when decision makers are faced with massive amounts of information, they are prone to information overload, which affects decision-making efficiency. Summary of the invention

[0004] In view of this, the embodiments of the present disclosure provide a method, device and equipment for constructing a decision prediction model based on visual information to solve the problems existing in the related art.

[0005] According to a first aspect of an embodiment of the present disclosure, a method for constructing a decision prediction model is provided, the method comprising: constructing an eye movement dataset, wherein the eye movement dataset comprises visual attributes reflected by different eye movement data in a decision-making process; obtaining multiple candidate models, and determining a target model from the multiple candidate models based on the accuracy rates and Kappa coefficients corresponding to the multiple candidate models; determining target hyperparameters corresponding to the target model using a ten-fold cross-validation method, and training the target model using the target hyperparameters and the eye movement dataset to obtain a decision prediction model.

[0006] A second aspect of an embodiment of the present disclosure provides a device for constructing a decision prediction model, which is applied to the method for constructing a decision prediction model as in the first aspect, and the device includes: a construction module, which is used to construct an eye movement data set, wherein the eye movement data set includes visual attributes reflected by different eye movement data in a decision-making process; an acquisition module, which is used to obtain multiple candidate models, and determine a target model from the multiple candidate models according to the accuracy and Kappa coefficient corresponding to the multiple candidate models; a model training module, which is used to determine the target hyperparameters corresponding to the target model using a ten-fold cross-validation method, and train the target model using the target hyperparameters and the eye movement data set to obtain a decision prediction model.

[0007] According to a third aspect of an embodiment of the present disclosure, a computer device is provided, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method for constructing the above-mentioned decision prediction model.

[0008] According to a fourth aspect of the embodiments of the present disclosure, a computer-readable storage medium is provided, on which a computer program / instruction is stored. When the computer program / instruction is executed by a processor, the steps of the method for constructing the decision prediction model are implemented.

[0009] According to a fifth aspect of the present disclosure, a computer program product is provided, and when the computer program / instructions are executed by a processor, the steps of the method for constructing the above-mentioned decision prediction model are implemented.

[0010] At least one of the above technical solutions adopted in the embodiments of the present disclosure can achieve the following beneficial effects: construct an eye movement data set, wherein the eye movement data set includes visual attributes reflected by different eye movement data in the decision-making process; obtain multiple candidate models, and determine the target model from the multiple candidate models according to the accuracy and Kappa coefficient corresponding to the multiple candidate models; use the ten-fold cross-validation method to determine the target hyperparameters corresponding to the target model, and use the target hyperparameters and the eye movement data set to train the target model to obtain a decision prediction model. Based on this, the eye movement data can be predicted by the target model, so that when faced with more eye movement data, it can be quickly analyzed and then generate corresponding decisions, thereby improving the efficiency of decision generation. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The above and other purposes, features and advantages of the present disclosure will become more apparent by describing the embodiments of the present disclosure in more detail in conjunction with the accompanying drawings. The accompanying drawings are used to provide a further understanding of the embodiments of the present disclosure and constitute a part of the specification. Together with the embodiments of the present disclosure, they are used to explain the present disclosure and do not constitute a limitation of the present disclosure. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0012] Figure 1 A flowchart of a method for constructing a decision prediction model provided by an exemplary embodiment of the present disclosure;

[0013] Figure 2 A flowchart of another method for constructing a decision prediction model provided by an exemplary embodiment of the present disclosure;

[0014] Figure 3 A schematic diagram of a structure of a device for constructing a decision prediction model provided by an exemplary embodiment of the present disclosure;

[0015] Figure 4 A schematic diagram of the structure of an electronic device provided for an exemplary embodiment of the present disclosure;

[0016] Figure 5 A schematic diagram of the structure of a computer system provided for an exemplary embodiment of the present disclosure;

[0017] Figure 6A schematic diagram of a computer program product provided for an exemplary embodiment of the present disclosure. DETAILED DESCRIPTION

[0018] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments described herein, which are instead provided for a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0019] It should be understood that the various steps described in the method embodiments of the present disclosure may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this respect.

[0020] The term "including" and its variations used in this document are open inclusions, that is, "including but not limited to". The term "based on" means "based at least in part on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one other embodiment"; the term "some embodiments" means "at least some embodiments". Relevant definitions of other terms will be given in the description below. It should be noted that the concepts of "first", "second", etc. mentioned in this disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0021] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0022] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0023] In recent years, with the rapid development of Internet technology, the user behaviors generated by users when using the Internet have become richer, and rich user behaviors will generate rich information, which provides a basis for decision-making in all aspects of user life. However, decision makers are prone to information overload when faced with massive amounts of information, which affects decision-making efficiency. It is worth noting that many user behaviors on the Internet involve time, such as shopping and browsing social media, which are closely related to life.

[0024] In actual applications, as digital technology and industrial systems mature, servers can automatically collect visual information from users when they perform these actions, and predict subsequent user behavior based on visual information, helping decision makers improve decision efficiency. Specifically, shopping websites will recommend products that consumers may be interested in based on browsing history and search habits, thereby reducing the time users spend on selecting products; social media platforms will push subsequent content based on the time users browse posts and their interactions, ensuring that the content users see is closely related to their interests and social activities.

[0025] Since the main source of information sampling in existing technologies is visual, visual information can be used to make effective decision predictions. However, few studies have systematically examined the predictive effects of visual information and its interactions on decision making. Machine learning algorithms can make up for this deficiency and can quantitatively compare the predictive performance of visual information on decision making and reveal the interactions between visual information.

[0026] As mentioned above, visual information can include eye movement data recorded by an eye tracker based on eye tracking technology to accurately measure the movement of the eyeball. This technology generally uses a pupil camera to collect information, and uses the principle of corneal reflection to identify and determine the distance change between the pupil center and the corneal reflection. The distance between these points is measured to obtain the coordinates of the eye gaze point to measure the eye movement information. Eye movement data information hides the user's attitude towards external stimuli and reflects the user's behavioral characteristics. The data volume is large and can reflect a variety of different attention characteristics at the same time, providing a good measurement for detecting how people pay attention, obtain and process information. With the development of science and technology, eye movement data can be collected not only through high-precision eye tracking instruments, but also directly through mobile devices such as mobile phones or computer cameras. Therefore, it has the characteristics of portability, non-invasiveness and low interference, and has many advantages in decision-making research.

[0027] Based on this, the embodiments of the present disclosure provide a method, device and equipment for constructing a decision prediction model based on visual information. A decision prediction model can be designed based on a neural network algorithm, and then the corresponding decision can be determined according to the user's current eye movement data. The use of the model can not only improve the efficiency of decision generation, but also quantitatively compare the prediction performance of eye movement indicators for decisions.

[0028] Figure 1 The following is a flow chart of a method for constructing a decision prediction model provided by an exemplary embodiment of the present disclosure. Figure 1 As shown, the specific steps include:

[0029] S101, obtaining eye movement data corresponding to the eye movement index, and constructing a data set according to the eye movement data.

[0030] In some embodiments, the disclosed embodiments select fixation frequency, single fixation duration, dwell time, transition and fixation location as eye movement indicators. Among them, fixation frequency is the number of fixations of the target divided by the total number of fixations; single fixation duration is the average fixation duration of the fixation point in each trial, reflecting the complexity of the data processing process; fixation duration is the length of time to fixate a certain point; transition is the transition of the eyes between different fixation points; fixation points record the movement trajectory of the eyes through eye tracking technology, and fixation point locations are the corresponding positions of each fixation point on the trajectory.

[0031] In practical applications, after determining the eye movement indicators, it is also necessary to determine the eye movement parameters included in each eye movement indicator. Specifically, for gaze frequency, five indicator parameters are mainly selected, including fixation points (Count-trial), fixation point distribution between options (Count-option), fixation point distribution between dimensions (Count-attribute), percentage of fixation (proportion-AOIacquired) and percentage of re-attention (proportion-AOIreacquired). The above dimensions can be set according to the situation. For example, when a user browses a shopping app and chooses to buy chocolate, faced with different choices of chocolate, the dimensions to be considered may be price, color, size and shipping.

[0032] The above fixation points are the total number of fixations on the information in the area of ​​interest in each trial, reflecting the frequency of individual fixation on the options; the distribution of fixations between options is to measure the relative number of fixations on the options. The absolute value of this measurement is limited to 1. A positive value indicates that the number of fixations on option A is greater, and a negative value indicates that the number of fixations on option B is greater. Option A and option B can be options that the current user can choose. For example, when a user browses a shopping app and chooses to buy chocolate, option A can be one type of chocolate and option B can be another type of chocolate. The specific calculation formula is:

[0033]

[0034] The distribution of fixations between dimensions is used to measure the relative number of fixations on the dimensions. The absolute value of this measurement is limited to 1. A positive value indicates that the number of fixations on dimension A is higher, and a negative value indicates that the number of fixations on dimension B is higher. The specific calculation formula is:

[0035]

[0036] The percentage of fixation is the proportion of the interest area that has been fixed at least once. The range of this measurement value is between 0 and 1. The closer it is to 1, the more comprehensive the individual's information processing in the interest area is. The specific calculation formula is:

[0037]

[0038] The percentage of re-gaze is the proportion of the interest area that has been gazed and re-gaze. The range of this measurement value is between 0 and 1. The closer it is to 1, the more interest areas are re-gaze. The specific calculation formula is:

[0039]

[0040] In practical applications, three indicator parameters are mainly selected for the duration of a single fixation point, including the average duration of a single fixation point (Duration-trial), the difference in the average duration of a single fixation point between options (Duration-option), and the difference in the average duration of a single fixation point between dimensions (Duration-attribute). Among them, the average duration of a single fixation point is the average duration of fixation of the fixation point in each trial, reflecting the complexity of the information processing process; the difference in the average duration of a single fixation point between options is the relative fixation duration of the fixation option, and the absolute limit of this measurement value is 1. A positive value indicates a longer fixation time on option A, and a negative value indicates a longer fixation time on option B. The specific calculation formula is:

[0041]

[0042] The difference in the average duration of a single fixation point between dimensions is the relative fixation duration of the fixation dimension. The absolute value of this measurement is limited to 1. A positive value indicates that the fixation time on dimension A is longer, and a negative value indicates that the fixation time on dimension B is longer. The specific calculation formula is:

[0043]

[0044] In practical applications, two indicator parameters are mainly selected for fixation duration, including attention allocation between options (SB-option) and attention allocation between dimensions (SB-attribute). Among them, attention allocation between options is the relative attention bias to the options. The absolute value of this measurement is limited to 1. A positive value indicates that the attention is more biased towards option A, and a negative value indicates that the attention is more biased towards option B. The specific calculation formula is:

[0045]

[0046] The attention allocation between dimensions is the relative attention bias to the dimensions. The absolute value of this measurement is limited to 1. A positive value indicates that the attention is more biased towards dimension A, and a negative value indicates that the attention is more biased towards dimension B. The specific calculation formula is:

[0047]

[0048] In practical applications, one indicator parameter is mainly selected for jumps, namely the distribution of option-based and dimension-based saccades (SM), and the specific calculation formula is:

[0049]

[0050] Among them, A represents the number of options; D represents the number of dimensions; r a represents the frequency of eye saccades based on the options; r d Represents the frequency of dimension-based saccades; N represents the total frequency of saccades; the larger the SM value, the higher the degree of dominance of option-based saccades in the frequency distribution; when the SM value is 0, it means that there is no dominant saccade pattern.

[0051] In practical applications, two indicator parameters are mainly selected for the gaze point position, including the first gaze point and the last gaze point. The last gaze point refers to whether the last gaze point falls on the preset target option when making a decision; the first gaze point refers to whether the first gaze point falls on the preset target option when making a decision.

[0052] In practical applications, after determining the parameters corresponding to each eye movement indicator, the data set required by the embodiment of the present disclosure is constructed by referring to the data in the existing database. Specifically, a decision-making construction data set available on the Internet can be selected, that is, visual information is widely collected, and a decision-making task based on preset parameters is tested on the Internet to obtain human factor data including eye movement data and behavioral data, and to test the performance of the automated prediction system. It should be understood that other indicators and corresponding indicator parameters can also be selected in practical applications, and no specific limitation is made here.

[0053] For example, for mobile terminal applications, when the server detects that the user's eyes are looking at the mobile phone screen area, the server determines a decision to wake up the screen based on the user's eye movement data, thereby lighting up the mobile phone screen; when the user is browsing shopping software, since the display interface corresponding to the shopping software may contain more than one product, and each product corresponds to a display area, the type of product that the user is interested in can be determined from multiple products based on the user's eye movement trajectory. Based on this, a decision can be made for the type of product that the user is interested in, and the decision can be to set 80% of the products in the display interface to products corresponding to that product type.

[0054] In practical applications, there are many examples of making decisions based on eye movement data. Therefore, the data in the existing database can be directly collected and referenced to construct the data set of the embodiment of the present disclosure. Specifically, the data corresponding to the five indicators mentioned above can be selected from the existing database to construct a data set, and the data set can include the eye movement data corresponding to the indicator parameters involved in the five eye movement indicators and the corresponding decisions. At the same time, the data set includes the visual attributes reflected by the different eye movement data in the decision-making process. It should be understood that the visual attributes are determined by the eye movement data corresponding to different eye movement parameters.

[0055] S102, determining a target model from multiple candidate models.

[0056] In some embodiments, a selection can be made from a plurality of candidate models for classification problems according to the performance of the model. Specifically, the candidate models of the disclosed embodiments include: generalized linear model, elastic network generalized linear model, random forest, regularized random forest and distributed gradient enhancement library (Extreme Gradient Boosting, XGBoost). Among them, the generalized linear model is a statistical learning method for describing and analyzing the relationship between dependent variables and independent variables. The model extends the traditional linear regression model, allowing the dependent variable to be not only a continuous numerical variable, but also a binary, multivariate, count or categorical variable; the elastic network generalized linear model is a generalized linear model fused with an elastic network, wherein the elastic network is a regularization technique that combines the characteristics of Lasso regression and ridge regression. By combining Lasso regression and ridge regression penalties, the elastic network effectively balances the trade-off between variable selection and coefficient shrinkage. This dual method can achieve a more robust model, especially in high-dimensional data sets that traditional methods may be difficult to handle; random forest is an ensemble learning algorithm consisting of one or more decision trees, which is usually used for classification and regression problems. Each decision tree performs classification based on its own feature selection and parameter settings, and the final result is determined by the classification result that appears most frequently in all decision trees; Regularization is a method of controlling the complexity of the model by adding a penalty term to the loss function, and regularized random forest is an algorithm that combines regularization and random forest; XGBoost is an optimized distributed gradient boosting library designed to be efficient, flexible, and portable. It implements machine learning algorithms under the framework of machine learning algorithms and can solve many data science problems quickly and accurately. It should be understood that other models can also be selected as candidate models in practical applications, and no specific restrictions are made here.

[0057] In practical applications, several influencing indicators such as accuracy, sensitivity, specificity, F1 score (F-measure) and Kappa statistics can be selected to evaluate the above candidate models, and the target model can be determined based on the evaluation results.

[0058] Among them, accuracy represents the percentage of correctly classified observations, which is the most important and most commonly used parameter for evaluating any classifier. The specific calculation formula is:

[0059]

[0060] Among them, Ac represents the accuracy; TP represents the positive samples predicted by the model as positive; TN represents the negative samples predicted by the model as negative; FP represents the negative samples predicted by the model as positive; FN represents the positive samples predicted by the model as negative.

[0061] Sensitivity represents the proportion of actual positive examples correctly identified by the model. The sensitivity component of actual positive values ​​to all expected positive values ​​is an important indicator for evaluating model performance. The specific calculation formula is:

[0062]

[0063] Among them, Sen represents the value corresponding to sensitivity.

[0064] Specificity indicates the proportion of actual negative examples correctly identified by the model, which is another important criterion for more accurately evaluating the prediction performance of the model. The specific calculation formula is:

[0065]

[0066] Wherein, Spe represents the numerical value corresponding to specificity.

[0067] The F1 score represents the weighted harmonic mean of the model's precision and recall, and is used to evaluate the overall performance of the model. The specific calculation formula is:

[0068]

[0069] Among them, Fm represents the value corresponding to the F1 score.

[0070] The Kappa statistic assesses the expected and observed consistency of qualitative features and is calculated as:

[0071]

[0072] Among them, K p represents the value corresponding to the Kappa statistic; P0 represents the accuracy of the prediction, which can also be understood as the consistency of the prediction; Pe Indicates accidental consistency.

[0073] Based on this, the candidate model with the highest accuracy and Kappa coefficient can be selected as the final target model.

[0074] S103, determining the optimal hyperparameters corresponding to the target model.

[0075] In some embodiments, a cross-validation method can be selected to determine the optimal hyperparameters corresponding to the target model.

[0076] Specifically, during the hyperparameter tuning process, the target model can be tested using the 10-fold cross validation method. In practical applications, the data set obtained by S101 can be randomly divided into 10 parts, one of which is taken as the validation set, and the remaining 9 parts are used to train the target model. Finally, the optimal hyperparameters are selected based on the average highest accuracy of the 10-fold.

[0077] S104, training the target model using the optimal hyperparameters and data set to obtain a decision prediction model.

[0078] In some embodiments, when the optimal hyperparameters are determined, the data in the original data set is divided into a training set and a test set in a ratio of 8:2. Based on this, the target model is trained to obtain a decision prediction model.

[0079] In practical applications, after the decision prediction model is successfully constructed, the Shapley value can also be calculated using the Shapley value calculation method to understand the contribution of each eye movement indicator to the output variable. Among them, the Shapley value is the value assigned to each feature in the sample. Shapley value was proposed by Shapley L.S in 1953 to solve the problem of conflicts caused by the distribution of benefits in the process of cooperation among multiple players. It belongs to the field of cooperative games. A major advantage of using the Shapley value is that the benefits are distributed according to the marginal contribution rate of the members to the alliance, that is, the benefits shared by the members are equal to the average value of the marginal benefits created by the member for the alliance he participates in.

[0080] Specifically, the Shapley value method can be used to calculate the Shapley value to explain the prediction logic of the eye movement index on the output variables of the decision prediction model. Explain that the decision prediction model g is a linear function of the feature attribute value, which is specifically expressed as:

[0081]

[0082] Among them, Φ0 represents the initial expectation of the model output for any observed eye movement indicator (base value), m represents the total number of eye movement indicators, and Φ iRepresents the eye movement index i belonging to the eye movement index, i represents the sequence number corresponding to the eye movement index, z ′ represents the input of the model, which is the simplified data of the original eye movement data. It should be understood that the eye movement indicators observed in the experiment correspond to z ′ i =1, or no eye movement indicator z was observed ′ i =0.

[0083] The classic SHAP value calculation formula is as follows:

[0084]

[0085] Among them, M represents the set of all eye movement indicators, mainly emphasizing a single eye movement indicator, S represents a subset of all eye movement indicators, and the SHAP value of eye movement indicator i represents the impact of adding eye movement indicator i on the model output. Two models are trained to measure the impact of eye movement indicator i. The first model f S (x S ) does not train the eye movement indicator i, another model f S∪{i} (x S∪{i} ) trained the eye movement index i, x S∪{i} is the eye movement index value entered for the eye movement index belonging to subset S and eye movement index i, x S is the eye movement index value input for subset S, and {i} is the set with only eye movement index i. For each possible subset Calculate the model output f S∪{i} (x S∪{i} )-f S (x S ) difference, and its weighted sum is the SHAP value of eye movement index i. The SHAP value not only reflects the contribution of the input eye movement index to the model prediction, but also reflects the direction of the influence on the model (positive or negative).

[0086] Since the eye movement index values ​​calculated in the above formula cannot reflect the pairwise interaction between eye movement indicators, the SHAP interaction value is proposed to capture the interaction effect. The above classic SHAP value is expanded to the SHAP interaction index, and the specific formula is as follows:

[0087]

[0088] Among them, j is the serial number of the eye movement index, i≠j, and, i,j(S)=f(S∪{i,j})-f(S∪{i})-f(S∪{j})+f(S), S∪{i} is the eye movement index belonging to subset S and eye movement index i, S∪{j} is the eye movement index belonging to subset S and eye movement index j, {i,j} is the set of only eye movement index i and eye movement index j, N is the set of all eye movement indicators, and the main emphasis is on the interaction between eye movement index i and eye movement index j.

[0089] According to the formula above, the SHAP interaction value Φ between the eye movement index i and the eye movement index j is calculated i,j is evenly distributed between pairs of eye movement indicators, so Φ i,j =Φ j,i , the total interaction between eye movement index i and eye movement index j is Φ i,j +Φ j,i Among them, Φ j,i Represents the SHAP interaction value between eye movement metric j and eye movement metric i.

[0090] The method for constructing a decision prediction model provided by the embodiments of the present disclosure may be executed by a terminal or by a chip applied to the terminal.

[0091] Exemplarily, the above-mentioned terminals may include one or more of mobile phones, tablet computers, wearable devices, vehicle-mounted devices, laptop computers, ultra-mobile personal computers (UMPCs), netbooks, personal digital assistants (PDAs), and wearable devices based on augmented reality (AR) and / or virtual reality (VR) technology, and may also include but not be limited to remote control devices, wearable devices, street lamps, home appliances and other smart terminals, which are not specifically limited in the embodiments of the present disclosure.

[0092] Figure 2 A flowchart of another method for constructing a decision prediction model provided by an exemplary embodiment of the present disclosure is shown below. Figure 2 As shown, specifically including:

[0093] S201, constructing an eye movement data set, wherein the eye movement data set includes visual attributes reflected by different eye movement data in a decision-making process.

[0094] It should be understood that eye movement data can be used to measure visual attributes, and visual attributes can specifically reflect the state of an object at a certain moment. For example, when a user is looking at a certain position on the mobile phone screen, it can be considered that during the gaze duration, the eye movement data corresponding to different eye movement parameters of the user remain unchanged during the gaze duration, and then the eye movement data corresponding to different eye movement parameters in this state can be used to determine the visual attributes of the user during this period, thereby helping to analyze the user's state at this time.

[0095] In some embodiments, a decision-making data set available on the Internet can be selected, that is, eye movement data can be widely collected, and a decision task with preset parameters can be tested on the Internet to obtain human factor data including eye movement data and behavioral data to test the performance of the automated prediction system. It should be understood that other indicators and corresponding indicator parameters can also be selected in actual applications, and no specific limitation is made here.

[0096] For example, for mobile terminal applications, when the server detects that the user's eyes are looking at the mobile phone screen area, the user's visual attributes are determined based on the user's eye movement data. The visual attributes reflect that the user is currently viewing the mobile phone screen, and then the decision is made to wake up the screen, thereby lighting up the mobile phone screen; when the user is browsing shopping software, since the display interface corresponding to the shopping software may contain more than one product, and each product corresponds to a display area, the user's visual attributes can be determined based on the user's eye movement trajectory. The visual attributes can reflect that the user has been browsing Category A products. At this time, the type of product that the user is interested in can be determined from multiple products, and then a decision can be made based on the type of product that the user is interested in. The decision can be to set 80% of the products in the display interface to products corresponding to that product type.

[0097] In actual applications, there are many examples of determining decisions through eye movement data. Therefore, the visual attributes reflected by different eye movement data in the existing database can be directly collected and referenced to construct the data set of the embodiment of the present disclosure. The visual attributes reflected by different eye movement data in the existing database and their corresponding decisions can be referenced to construct the data set of the embodiment of the present disclosure. The database can also be directly constructed based on the eye movement data corresponding to different decisions.

[0098] S202, obtaining multiple candidate models, and determining a target model from the multiple candidate models according to the accuracy and Kappa coefficients corresponding to the multiple candidate models.

[0099] In some embodiments, a plurality of candidate models for classification problems can be selected based on the performance of the model. Specifically, the candidate models of the disclosed embodiments include: generalized linear model, elastic net generalized linear model, random forest, regularized random forest and distributed gradient boosting library (Extreme Gradient Boosting, XGBoost).

[0100] In practical applications, several influencing indicators such as accuracy, sensitivity, specificity, F1 score (F-measure) and Kappa statistics can be selected to evaluate the above candidate models, and the target model can be determined based on the evaluation results.

[0101] Since accuracy represents the percentage of correctly classified observations, which is the most important and most commonly used parameter for evaluating any classifier, and Kappa statistic evaluates the expected and observed consistency of qualitative features, the embodiments of the present disclosure may mainly consider accuracy and Kappa statistic.

[0102] Specifically, the candidate model with the highest accuracy and Kappa coefficient can be selected as the target model, the candidate model with the highest accuracy can be selected as the target model, and the candidate model with the highest Kappa coefficient can be selected as the target model.

[0103] S203, using a ten-fold cross-validation method to determine the target hyperparameters corresponding to the target model, and using the target hyperparameters and the eye movement dataset to train the target model to obtain a decision prediction model.

[0104] In some embodiments, the data set obtained in S201 can be randomly divided into ten parts, one of which is taken as a validation set in turn, and the remaining nine parts are used to train the target model. Finally, the optimal hyperparameters are selected based on the ten-fold average highest accuracy.

[0105] By evaluating and improving the hyperparameters of the target model through ten-fold cross validation, the number of training samples can be maximized, thereby achieving more accurate model performance evaluation and minimizing the bias problem caused by uneven data distribution. Secondly, the impact of multicollinearity can be accurately reduced and nonlinear interactions can be discussed.

[0106] Based on this, when the optimal hyperparameters are determined, the data in the original data set is divided into a training set and a test set in a ratio of 8:2. Based on this, the target model is trained to obtain a decision prediction model.

[0107] In summary, the disclosed embodiment constructs an eye movement data set, wherein the eye movement data set includes visual attributes reflected by different eye movement data in the decision-making process; obtains multiple candidate models, and determines a target model from the multiple candidate models according to the accuracy and Kappa coefficient corresponding to the multiple candidate models; uses a ten-fold cross-validation method to determine the target hyperparameters corresponding to the target model, and uses the target hyperparameters and the eye movement data set to train the target model to obtain a decision prediction model. Based on this, the eye movement data can be predicted by the target model, so that when faced with a large amount of eye movement data, it can be quickly analyzed and the corresponding decision can be generated.

[0108] In some embodiments, constructing an eye movement dataset includes: determining an eye movement indicator and determining eye movement parameters included in the eye movement indicator; determining eye movement data corresponding to the eye movement parameters during a decision-making process; and constructing the eye movement dataset based on visual attributes reflected by the eye movement data.

[0109] Specifically, the eye movement indicators may include fixation frequencies, fixation duration, dwell time, transition, and fixation location as eye movement indicators.

[0110] After determining the eye movement indicators, it is also necessary to determine the eye movement parameters included in each eye movement indicator. Specifically, for gaze frequency, five indicator parameters were mainly selected, including fixation points (Count-trial), fixation point distribution between options (Count-option), fixation point distribution between dimensions (Count-attribute), percentage of fixation (Proportion-AOIacquired) and percentage of return gaze (proportion-AOIreacquired).

[0111] For the duration of a single fixation point, three indicator parameters were mainly selected, including the average duration of a single fixation point (Duration-trial), the difference in the average duration of a single fixation point between options (Duration-option), and the difference in the average duration of a single fixation point between dimensions (Duration-attribute).

[0112] In practical applications, two indicator parameters are mainly selected for fixation duration, including attention allocation between options (SB-option) and attention allocation between dimensions (SB-attribute).

[0113] In practical applications, one indicator parameter is mainly selected for jumps, namely the distribution of option-based and dimension-based saccades (SM).

[0114] For the fixation point position, two indicator parameters were mainly selected, including the first fixation point and the last fixation point.

[0115] In practical applications, after determining the parameters corresponding to each eye movement indicator, the data set required by the embodiment of the present disclosure is constructed by referring to the data in the existing database. Specifically, a decision-making construction data set available on the Internet can be selected, that is, visual information is widely collected, and a decision-making task based on preset parameters is tested on the Internet to obtain human factor data including eye movement data and behavioral data, and to test the performance of the automated prediction system. It should be understood that other indicators and corresponding indicator parameters can also be selected in practical applications, and no specific limitation is made here.

[0116] For example, for mobile terminal applications, when the server detects that the user's eyes are looking at the mobile phone screen area, the user's visual attributes are determined based on the user's eye movement data. The visual attributes reflect that the user is currently viewing the mobile phone screen, and then the decision is made to wake up the screen, thereby lighting up the mobile phone screen; when the user is browsing shopping software, since the display interface corresponding to the shopping software may contain more than one product, and each product corresponds to a display area, the user's visual attributes can be determined based on the user's eye movement trajectory. The visual attributes can reflect that the user has been browsing Category A products. At this time, the type of product that the user is interested in can be determined from multiple products, and then a decision can be made based on the type of product that the user is interested in. The decision can be to set 80% of the products in the display interface to products corresponding to that product type.

[0117] In practical applications, there are many examples of determining decisions through eye movement data. Therefore, the visual attributes reflected by different eye movement data in the existing database can be directly collected to construct the data set of the embodiment of the present disclosure. The visual attributes reflected by different eye movement data in the existing database and their corresponding decisions can also be used to construct the data set of the embodiment of the present disclosure. The database can also be directly constructed based on the eye movement data corresponding to different decisions. Among them, in the process of constructing the data set of the embodiment of the present disclosure based on the data in the existing database, it is also necessary to clear the invalid data in the database, and integrate the composite features of the remaining data after clearing the invalid data. After integrating the composite features, it is also necessary to standardize the data, and then the data set of the embodiment of the present disclosure can be constructed based on the standardized data. The above-mentioned integration of composite features refers to the creation of new features from the original data, and these new features are obtained by combining or transforming multiple original features. This method aims to enhance the expressiveness of the model so that it can better capture patterns and relationships in the data. Integrating composite features can improve the predictive ability of the model, especially when the original features cannot fully express the complexity of the data.

[0118] When there are multiple eye movement indicators, the Shapley value calculation method is used to determine the contribution value corresponding to each eye movement indicator. The contribution value is used to determine the contribution of each eye movement indicator to the output result of the decision prediction model, which can explain the reason for the model prediction and thus improve the transparency and interpretability of the model.

[0119] In practical applications, after the decision prediction model is successfully constructed, the Shapley value can also be calculated using the Shapley value calculation method to understand the contribution of each eye movement indicator to the output variable. Among them, the Shapley value is the value assigned to each feature in the sample. Shapley value was proposed by Shapley L.S in 1953 to solve the problem of conflicts caused by the distribution of benefits in the process of cooperation among multiple players. It belongs to the field of cooperative games. A major advantage of using the Shapley value is that the benefits are distributed according to the marginal contribution rate of the members to the alliance, that is, the benefits shared by the members are equal to the average value of the marginal benefits created by the member for the alliance he participates in.

[0120] Specifically, the Shapley value method can be used to calculate the Shapley value to explain the prediction logic of the eye movement index on the output variables of the decision prediction model. Explain that the decision prediction model g is a linear function of the feature attribute value, which is specifically expressed as:

[0121]

[0122] Among them, Φ0 represents the initial expectation of the model output for any observed eye movement indicator (base value), m represents the total number of eye movement indicators, and Φ i represents the eye movement index attribution of feature i, z ′ represents the input of the model, which is the simplified data of the original eye movement data. It should be understood that the eye movement indicators observed in the experiment correspond to z ′ i =1, or no eye movement indicator z was observed ′ i =0.

[0123] The classic SHAP value calculation formula is as follows:

[0124]

[0125] Among them, M represents the set of all eye movement indicators, mainly emphasizing a single eye movement indicator, S represents a subset of all eye movement indicators, and the SHAP value of eye movement indicator i represents the impact of adding eye movement indicator i on the model output. Two models are trained to measure the impact of eye movement indicator i. The first model f S (x S ) does not train the eye movement indicator i, another model f S∪{i} (x S∪{i} ) trained the eye movement index i, x S∪{i} is the eye movement index value entered for the eye movement index belonging to subset S and eye movement index i, x S is the eye movement index value input for subset S, and {i} is the set with only eye movement index i. For each possible subset Calculate the model output f S∪{i} (x S∪{i} )-f S (x S ) difference, and its weighted sum is the SHAP value of eye movement index i. The SHAP value not only reflects the contribution of the input eye movement index to the model prediction, but also reflects the direction of the influence on the model (positive or negative).

[0126] Since the eye movement index values ​​calculated in the above formula cannot reflect the pairwise interaction between eye movement indicators, the SHAP interaction value is proposed to capture the interaction effect. The above classic SHAP value is expanded to the SHAP interaction index, and the specific formula is as follows:

[0127]

[0128] Among them, j is the serial number of the eye movement index, i≠j, and, i,j(S)=f(S∪{i,j})-f(S∪{i})-f(S∪{j})+f(S), S∪{i} is the eye movement index belonging to subset S and eye movement index i, S∪{j} is the eye movement index belonging to subset S and eye movement index j, {i,j} is the set of only eye movement index i and eye movement index j, N is the set of all eye movement indicators, and the main emphasis is on the interaction between eye movement index i and eye movement index j.

[0129] According to the formula above, the SHAP interaction value Φ between the eye movement index i and the eye movement index j is calculated i,j is evenly distributed between pairs of eye movement indicators, so Φ i,j =Φ j,i , the total interaction between eye movement index i and eye movement index j is Φ i,j +Φ j,i Among them, Φ j,i Represents the SHAP interaction value between eye movement metric j and eye movement metric i.

[0130] Based on this, the embodiments of the present disclosure can quantitatively compare the prediction performance of eye movement indicators by combining machine learning technology and SHAPLY-based model interpretation, effectively identify features that are closely related to decision results, and establish a decision prediction method based on eye movement data.

[0131] In some embodiments, a ten-fold cross-validation method is used to determine target hyperparameters corresponding to the target model, including: dividing the eye movement data set into a training set and a validation set; inputting the training set and the validation set into the target model, and training using the ten-fold cross-validation method, and determining the target hyperparameters based on the training results.

[0132] Specifically, in the process of hyperparameter tuning, the 10-fold cross validation method can be used to test the model. In practical applications, the data set can be randomly divided into 10 parts, one of which is taken as the validation set, and the remaining 9 parts are used to train the target model. Finally, the optimal hyperparameters are selected based on the average highest accuracy of the 10-fold cross validation.

[0133] In some embodiments, a target model is determined from multiple candidate models based on the accuracy and Kappa coefficients corresponding to multiple candidate models, including: determining the candidate model with the highest accuracy from multiple candidate models as the target model; and / or determining the candidate model with the highest Kappa coefficient from multiple candidate models as the target model.

[0134] In practical applications, several influencing indicators such as accuracy, sensitivity, specificity, F1 score (F-measure) and Kappa statistics can be selected to evaluate the above candidate models, and the target model can be determined based on the evaluation results.

[0135] Among them, accuracy represents the percentage of correctly classified observations, which is the most important and most commonly used parameter for evaluating any classifier. The specific calculation formula is:

[0136]

[0137] Among them, Ac represents the accuracy; TP represents the positive samples predicted by the model as positive; TN represents the negative samples predicted by the model as negative; FP represents the negative samples predicted by the model as positive; FN represents the positive samples predicted by the model as negative.

[0138] Sensitivity represents the proportion of actual positive examples correctly identified by the model. The sensitivity component of actual positive values ​​to all expected positive values ​​is an important indicator for evaluating model performance. The specific calculation formula is:

[0139]

[0140] Among them, Sen represents the value corresponding to sensitivity.

[0141] Specificity indicates the proportion of actual negative examples correctly identified by the model, which is another important criterion for more accurately evaluating the prediction performance of the model. The specific calculation formula is:

[0142]

[0143] Wherein, Spe represents the numerical value corresponding to specificity.

[0144] The F1 score represents the weighted harmonic mean of the model's precision and recall, and is used to evaluate the overall performance of the model. The specific calculation formula is:

[0145]

[0146] Among them, Fm represents the value corresponding to the F1 score.

[0147] The Kappa statistic assesses the expected and observed consistency of qualitative features and is calculated as:

[0148]

[0149] Among them, K p represents the value corresponding to the Kappa statistic; P0 represents the accuracy of the prediction, which can also be understood as the consistency of the prediction; Pe Indicates accidental consistency.

[0150] Since accuracy represents the percentage of correctly classified observations, which is the most important and most commonly used parameter for evaluating any classifier, and Kappa statistic evaluates the expected and observed consistency of qualitative features, the embodiments of the present disclosure may mainly consider accuracy and Kappa statistic.

[0151] Specifically, the candidate model with the highest accuracy and Kappa coefficient can be selected as the target model, the candidate model with the highest accuracy can be selected as the target model, and the candidate model with the highest Kappa coefficient can be selected as the target model.

[0152] In summary, the embodiments of the present disclosure firstly evaluate and improve the hyperparameters of the target model through ten-fold cross validation, which can maximize the number of training samples, thereby achieving more accurate target model performance evaluation and minimizing the bias problem caused by uneven data distribution. Secondly, the impact of multicollinearity can be accurately reduced, and nonlinear interactions can be discussed. Finally, by combining machine learning technology and SHAPLY-based model interpretation, the prediction performance of eye movement indicators is quantitatively compared, and features closely related to decision results are effectively identified. A decision prediction method for eye movement data is established, which can ensure the accuracy of decision generation while ensuring the efficiency of decision generation.

[0153] The above mainly introduces the solution provided by the embodiment of the present disclosure from the perspective of the server. It is understandable that, in order to implement the above functions, the server includes hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should easily realize that, in combination with the units and algorithm steps of each example described in the embodiments disclosed herein, the present disclosure can be implemented in the form of hardware or a combination of hardware and computer software. Whether a function is executed in the form of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present disclosure.

[0154] The embodiments of the present disclosure may divide the server into functional units according to the above method examples. For example, each functional module may be divided corresponding to each function, or two or more functions may be integrated into one management module. The above integrated modules may be implemented in the form of hardware or in the form of software functional modules. It should be noted that the division of modules in the embodiments of the present disclosure is schematic and is only a logical functional division. There may be other division methods in actual implementation.

[0155] In the case of dividing each functional module according to each function, an exemplary embodiment of the present disclosure provides a device for constructing a decision prediction model, and the device for constructing a decision prediction model may be a server or a chip applied to a server. Figure 3 The following is a schematic diagram of a device for constructing a decision prediction model provided by an exemplary embodiment of the present disclosure. Figure 3 As shown, the decision prediction model construction device 300 includes:

[0156] A construction module 301 is used to construct an eye movement data set, wherein the eye movement data set includes visual attributes reflected by different eye movement data in a decision-making process;

[0157] An acquisition module 302 is used to acquire multiple candidate models and determine a target model from the multiple candidate models according to the accuracy and Kappa coefficients corresponding to the multiple candidate models;

[0158] The model training module 303 is used to determine the target hyperparameters corresponding to the target model using a ten-fold cross-validation method, and train the target model using the target hyperparameters and the eye movement dataset to obtain a decision prediction model.

[0159] In an optional manner, the construction module 301 is also used to determine eye movement indicators and eye movement parameters included in the eye movement indicators; determine eye movement data corresponding to the eye movement parameters during the decision-making process; and construct the eye movement data set based on the visual attributes reflected by the eye movement data.

[0160] In an optional manner, the decision prediction model construction device 300 also includes a determination module 304 for determining, when there are multiple eye movement indicators, the contribution value corresponding to each eye movement indicator by using the Shapley value calculation method, wherein the contribution value is used to determine the contribution of each eye movement indicator to the output result of the decision prediction model.

[0161] In an optional manner, the eye movement indicators include gaze frequency, single gaze point duration, gaze duration, jumps and gaze point positions.

[0162] In an optional manner, the model training module 303 is also used to divide the eye movement dataset into a training set and a validation set; input the training set and the validation set into the target model, and use a ten-fold cross-validation method for training, and determine the target hyperparameters based on the training results.

[0163] In an optional manner, the acquisition module 302 is further used to determine the candidate model with the highest accuracy from multiple candidate models as the target model; and / or determine the candidate model with the highest Kappa coefficient from multiple candidate models as the target model.

[0164] An embodiment of the present disclosure also provides an electronic device, comprising: at least one processor; a memory for storing instructions executable by at least one processor; wherein the at least one processor is used to execute instructions to implement the steps of the above method disclosed in the embodiment of the present disclosure.

[0165] Figure 4 The structure diagram of an electronic device provided by an exemplary embodiment of the present disclosure is shown in FIG. Figure 4 As shown, the electronic device 400 includes at least one processor 401 and a memory 402 coupled to the processor 401 , and the processor 401 can execute corresponding steps in the above method disclosed in the embodiment of the present disclosure.

[0166] The processor 401 may also be referred to as a central processing unit (CPU), which may be an integrated circuit chip having signal processing capabilities. Each step in the method disclosed in the embodiment of the present disclosure may be completed by an integrated logic circuit of hardware in the processor 401 or by instructions in the form of software. The processor 401 may be a general-purpose processor, a digital signal processor (DSP), an ASIC, a field-programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in the embodiment of the present disclosure may be directly embodied as being executed by a hardware decoding processor, or may be executed by a combination of hardware and software modules in a decoding processor. The software module may be located in a memory 402, such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, or other mature storage media in the art. The processor 401 reads the information in the memory 402 and completes the steps of the method in combination with its hardware.

[0167] In addition, when various operations / processes according to the present disclosure are implemented by software and / or firmware, they can be transmitted from a storage medium or a network to a computer system having a dedicated hardware structure, for example, Figure 5 The computer system 500 shown installs the programs constituting the software, and when the various programs are installed, the computer system can execute various functions, including the functions described above. Figure 5 A schematic diagram of the structure of a computer system provided by an exemplary embodiment of the present disclosure.

[0168] Computer system 500 is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Electronic devices may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0169] like Figure 5 As shown, the computer system 500 includes a computing unit 501, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 502 or a computer program loaded from a storage unit 508 into a random access memory (RAM) 503. In the RAM 503, various programs and data required for the operation of the computer system 500 can also be stored. The computing unit 501, the ROM 502, and the RAM 503 are connected to each other via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0170] A plurality of components in the computer system 500 are connected to the I / O interface 505, including: an input unit 506, an output unit 507, a storage unit 508, and a communication unit 509. The input unit 506 may be any type of device capable of inputting information to the computer system 500, and the input unit 506 may receive input digital or character information, and generate key signal inputs related to user settings and / or function control of the electronic device. The output unit 507 may be any type of device capable of presenting information, and may include, but is not limited to, a display, a speaker, a video / audio output terminal, a vibrator, and / or a printer. The storage unit 508 may include, but is not limited to, a disk, an optical disk. The communication unit 509 allows the computer system 500 to exchange information / data with other devices over a network such as the Internet, and may include, but is not limited to, a modem, a network card, an infrared communication device, a wireless communication transceiver, and / or a chipset, for example, a Bluetooth™ device, a WiFi device, a WiMax device, a cellular communication device, and / or the like.

[0171] The computing unit 501 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 501 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 501 performs the various methods and processes described above. For example, in some embodiments, the above methods disclosed in the embodiments of the present disclosure may be implemented as a computer software program, which is tangibly included in a machine-readable medium, such as a storage unit 508. In some embodiments, part or all of the computer program may be loaded and / or installed on an electronic device via a ROM 502 and / or a communication unit 509. In some embodiments, the computing unit 501 may be configured to perform the above methods disclosed in the embodiments of the present disclosure in any other appropriate manner (e.g., by means of firmware).

[0172] The embodiment of the present disclosure also provides a computer-readable storage medium, wherein when the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the above method disclosed in the embodiment of the present disclosure.

[0173] The computer-readable storage medium in the disclosed embodiments may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. The computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or equipment, or any suitable combination of the foregoing. More specifically, the computer-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0174] The computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.

[0175] Figure 6 Schematic diagram of a computer program product provided by an exemplary embodiment of the present disclosure. Figure 6 As shown, the computer program product 600 includes a computer program 601, wherein the computer program 601 implements the above method disclosed in the embodiment of the present disclosure when executed by a processor.

[0176] In embodiments of the present disclosure, computer program codes for performing the operations of the present disclosure may be written in one or more programming languages ​​or combinations thereof, including but not limited to object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" language or similar programming languages. The program code may be executed entirely on a user's computer, partially on a user's computer, as an independent software package, partially on a user's computer, partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer.

[0177] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0178] The modules, components or units involved in the embodiments described in the present disclosure may be implemented by software or hardware, wherein the names of the modules, components or units do not, in some cases, limit the modules, components or units themselves.

[0179] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chip (SOCs), complex programmable logic devices (CPLDs), and the like.

[0180] The above descriptions are only some embodiments of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present disclosure is not limited to the technical solutions formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, a technical solution formed by replacing the above features with the technical features with similar functions disclosed in the present disclosure (but not limited to).

[0181] Although some specific embodiments of the present disclosure have been described in detail by way of example, it should be understood by those skilled in the art that the above examples are for illustration only and are not intended to limit the scope of the present disclosure. It should be understood by those skilled in the art that the above embodiments may be modified without departing from the scope and spirit of the present disclosure. The scope of the present disclosure is defined by the appended claims.

Claims

1. A method for constructing a decision prediction model, characterized in that: include: Constructing an eye movement data set, wherein the eye movement data set includes visual attributes reflected by different eye movement data in a decision-making process; Acquire multiple candidate models, and determine a target model from the multiple candidate models according to the accuracy and Kappa coefficients corresponding to the multiple candidate models; The target hyperparameters corresponding to the target model are determined using a ten-fold cross-validation method, and the target model is trained using the target hyperparameters and the eye movement dataset to obtain a decision prediction model.

2. The method according to claim 1, characterized in that The step of constructing an eye movement dataset comprises: Determine an eye movement indicator, and determine the eye movement parameters included in the eye movement indicator; Determining eye movement data corresponding to the eye movement parameters during the decision-making process; The eye movement data set is constructed according to the visual attributes reflected by the eye movement data.

3. The method according to claim 2, characterized in that The method further comprises: In the case where there are multiple eye movement indicators, the contribution value corresponding to each eye movement indicator is determined respectively using the Shapley value calculation method, wherein the contribution value is used to determine the contribution of each eye movement indicator to the output result of the decision prediction model.

4. The method according to claim 3, characterized in that The eye movement indicators include fixation frequency, single fixation time, fixation time, jump and fixation position.

5. The method according to claim 1, characterized in that The method of using a ten-fold cross validation method to determine the target hyperparameters corresponding to the target model includes: Dividing the eye movement dataset into a training set and a validation set; The training set and the validation set are input into the target model, and training is performed using a ten-fold cross validation method, and the target hyperparameters are determined according to the training results.

6. The method according to claim 1, characterized in that Determining a target model from a plurality of candidate models according to the accuracy rates and Kappa coefficients corresponding to the plurality of candidate models comprises: Determine the candidate model with the highest accuracy from multiple candidate models as the target model; And / or determine the candidate model with the highest Kappa coefficient from multiple candidate models as the target model.

7. A device for constructing a decision prediction model, characterized in that: include: A construction module, used to construct an eye movement data set, wherein the eye movement data set includes visual attributes reflected by different eye movement data in a decision-making process; <24P07701CN> An acquisition module, used to acquire multiple candidate models, and determine a target model from the multiple candidate models according to the accuracy and Kappa coefficients corresponding to the multiple candidate models; The model training module is used to determine the target hyperparameters corresponding to the target model using a ten-fold cross-validation method, and train the target model using the target hyperparameters and the eye movement dataset to obtain a decision prediction model.

8. A computer device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method of claim 1.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instructions are executed by a processor, the steps of the method according to claim 1 are implemented.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method according to claim 1 are implemented.

Images