Calibration method, device and equipment for thermodynamic diagram

Abstract

The embodiment of the invention discloses a thermodynamic diagram calibration method, device and equipment. The method comprises the steps that a current behavior track and a complete behavior track which are located in the same page session are acquired; the current behavior track is a sequence formed by user behavior data until the preset burying point reporting moment; determining adjacent subsequent behaviors of the current behavior track based on the complete behavior track of the user side; determining the state probability of each page state through the end behavior of the current behavior track and the adjacent subsequent behavior, and calibrating the page state corresponding to the current behavior track; and generating a page thermodynamic diagram according to the calibrated page state and the user behavior data.

Description

Technical Field

[0001] This specification relates to the field of data visualization technology, and in particular to a method, apparatus, and device for calibrating a heatmap. Background Technology

[0002] Currently, to statistically analyze user interactions on web pages, heatmaps are typically created to visually represent the page's activity. However, with increasing interface complexity, modern complex pages containing multimodal interactive components such as pop-ups and drawers can have completely different functions and user intentions depending on the screen coordinates and their states. Without identifying page states, heatmaps cannot accurately represent user intent.

[0003] In traditional methods, all user actions from entering the page to the current moment are recorded sequentially, forming a behavior sequence. This sequence is then compared with a predefined page state machine that describes page state transitions to identify the page's current state.

[0004] As the complexity of interface interactions increases, the state of in-app pages becomes more complex, typically involving two or more modal interaction components. This complexity significantly increases the difficulty of constructing state machines and introduces unexpected anomalies. Consequently, relying solely on the current behavior sequence combined with the page state machine is insufficient to accurately identify the page's state, resulting in random errors in the generated heatmap.

[0005] Therefore, a way to calibrate page status is needed to improve the accuracy of heatmaps. Summary of the Invention

[0006] This specification provides one or more embodiments of a heatmap calibration method, apparatus, and device to address the following technical problem: the need for a method to calibrate page states to improve the accuracy of heatmaps.

[0007] To solve the above-mentioned technical problems, one or more embodiments of this specification are implemented as follows: This specification provides a method for calibrating a heatmap through one or more embodiments, including: Obtain the current behavior trajectory and complete behavior trajectory of the session on the same page; the current behavior trajectory is a sequence of user behavior data up to the time of the preset data entry point reporting. Based on the complete behavioral trajectory of the user terminal, determine the adjacent subsequent behaviors of the current behavioral trajectory; By combining the final behavior of the current behavior trajectory with the adjacent subsequent behavior, the state probability of each page state is determined, and the page state corresponding to the current behavior trajectory is calibrated. A page heatmap is generated based on the calibrated page status and the user behavior data.

[0008] This specification provides a heatmap calibration device according to one or more embodiments, the device comprising: The acquisition module is used to acquire the current behavior trajectory and the complete behavior trajectory of a session on the same page; the current behavior trajectory is a sequence of user behavior data based on the time of reporting the preset embedded points; The determination module is used to determine the adjacent subsequent behaviors of the current behavior trajectory based on the complete behavior trajectory of the user terminal; The calibration module is used to determine the state probability of each page state by the end behavior of the current behavior trajectory and the adjacent subsequent behavior, and to calibrate the page state corresponding to the current behavior trajectory. The generation module is used to generate a page heatmap based on the calibrated page status and the user behavior data.

[0009] This specification provides one or more embodiments of a heatmap calibration device, comprising: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enable the at least one processor to: Obtain the current behavior trajectory and complete behavior trajectory of the session on the same page; the current behavior trajectory is a sequence of user behavior data up to the time of the preset data entry point reporting. Based on the complete behavioral trajectory of the user terminal, determine the adjacent subsequent behaviors of the current behavioral trajectory; By combining the final behavior of the current behavior trajectory with the adjacent subsequent behavior, the state probability of each page state is determined, and the page state corresponding to the current behavior trajectory is calibrated. A page heatmap is generated based on the calibrated page status and the user behavior data.

[0010] The at least one technical solution adopted in the embodiments of this specification can achieve the following beneficial effects: By backtracking the complete behavior trajectory, the adjacent subsequent behaviors of the current behavior trajectory are obtained, and the state probability of the page state is inferred based on the adjacent subsequent behaviors, thereby realizing the calibration of the page state and solving the problem of heatmap error caused by abnormality or uncertain initialization state in complex state machine scenarios. After introducing the complete behavior trajectory, the adjacent subsequent behaviors of the current behavior trajectory in the complete behavior trajectory are determined, realizing the backtracking calibration of user behavior data and avoiding misjudgment caused by incomplete information. Based on the state probability of each page state, the page state corresponding to the current behavior trajectory is calibrated, solving the problem of multiple possible states caused by abnormality or initialization ambiguity and realizing error compensation for page state uncertainty. Based on the accurate page state after calibration, user behavior data is classified and heatmap is rendered, so that the generated heatmap can accurately reflect user behavior under different page states. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments or prior art of this specification, the drawings used in the description of the embodiments or prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings: Figure 1 A schematic flowchart of a heatmap calibration method provided for embodiments of this specification; Figure 2 A schematic diagram of user behavior data in a page session under an application scenario, provided for an embodiment of this specification; Figure 3 A schematic diagram of an existing page state machine in an application scenario provided for the embodiments of this specification; Figure 4 A flowchart illustrating a process for calibrating the page state corresponding to the current behavior trajectory, provided as an embodiment of this specification. Figure 5 A schematic diagram of a preset page state machine in an application scenario provided for the embodiments of this specification; Figure 6 A schematic diagram of a page state calibration process in an application scenario, provided for the embodiments of this specification; Figure 7 This is a schematic diagram illustrating an application scenario where the page state cannot be calibrated, provided as an embodiment of this specification. Figure 8 A schematic diagram of a composite page state calibration process in an application scenario, provided for the embodiments of this specification; Figure 9A schematic diagram of a page heatmap in an application scenario provided by an embodiment of this specification; Figure 10 A schematic diagram of a thermal map calibration device provided for embodiments of this specification; Figure 11 This is a schematic diagram of a thermal mapping calibration device provided as an embodiment of this specification. Detailed Implementation

[0012] This specification provides a method, apparatus, and device for calibrating a heat map.

[0013] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0014] In the context of the background technology, heatmaps are a key technology in front-end user behavior analysis used to visualize user behaviors on web pages, such as clicks, swipes, and long presses. They use the warmth and shade of colors to visually represent the density or intensity distribution of user behavior data in a two-dimensional space. Currently, to statistically analyze user actions on web pages, heatmaps are typically created to visually demonstrate the page's interaction status. However, with the increasing complexity of user interfaces, in modern complex pages containing multimodal interactive components such as pop-ups and drawers, the same screen coordinates in different states carry completely different functions and user intentions. Without identifying page states, heatmaps cannot accurately represent user intentions.

[0015] To be more intuitive, provide examples of the solutions that the applicant has tried, and the specific issues they each involve.

[0016] The first type of approach, a traditional one, records all user actions from the moment they enter the page until the current moment, forming a sequence of behaviors. This sequence is then compared to a predefined page state machine that describes page state transitions to identify the page's current state.

[0017] As the complexity of interface interactions increases, such as in a certain activity scenario where the state of the in-app page is complex and there are two or more modal interaction components (modal interaction refers to the current interface forcing the user to focus on a specific interaction layer, during which the user cannot interact with other parts of the main interface until the modal layer is closed or the operation is completed, such as pop-ups, drawers, animations, etc.), the complexity of the state makes the construction of the state machine much more difficult, and there may be unexpected exceptions such as interface exceptions that cause state switching failures. The above situations will lead to the inability to accurately identify the state of the page by only combining the current behavior sequence with the page state machine, resulting in a certain degree of random error in the generated heatmap.

[0018] The second type of approach is state prediction based on machine learning. It trains a machine learning model by training historical data, and then uses the machine learning model and decision tree to predict the page state corresponding to the user behavior sequence.

[0019] This approach requires a large amount of labeled data to train the model, making it difficult to apply to small-scale operations. Furthermore, the model's predictions are difficult to interpret intuitively, affecting the reliability of the analysis. In addition, model inference is time-consuming and cannot calibrate heatmaps in real time.

[0020] The third type is rule engine-driven state calibration, which uses preset business rules such as regular expressions and state transition matrices to match user behavior sequences and calibrate the page state.

[0021] This approach requires rule-based calibration, necessitating frequent rule updates when business logic changes, resulting in high maintenance costs and an inability to handle unexpected states caused by interface anomalies. Furthermore, when the same behavior sequence corresponds to multiple states, rule conflicts cannot be resolved.

[0022] To address the aforementioned issues, this application proposes a heatmap error calibration method based on state uncertainty compensation. Through complete behavior trajectory backtracking and probabilistic statistical auxiliary calibration, it resolves heatmap errors caused by interface anomalies, programming language (JavaScript, JS) anomalies, or uncertain initialization states in complex state machine scenarios. In particular, by utilizing the immediately following subsequent behavior in the complete trajectory, the starting page state at the time of the final behavior can be uniquely or with high probability deduced from the state machine. This solves the problem of existing methods relying on the current behavior sequence and effectively handles interface anomalies or uncertain initialization states. Based on this overall approach, the solution of this application will be further described in detail below.

[0023] Figure 1This document provides a flowchart illustrating a heatmap calibration method according to one or more embodiments. This method can be applied to various business areas requiring analysis of user actions on web pages, such as internet products, enterprise applications, internet finance, and e-commerce live streaming. The process can be executed by computing devices relevant to the specific area (e.g., servers or clients corresponding to internet product businesses). Certain input parameters or intermediate results within the process can be manually adjusted to improve accuracy.

[0024] Figure 1 The process may include the following steps: S102: Obtain the current behavior trajectory and complete behavior trajectory of the session on the same page; the current behavior trajectory is a sequence of user behavior data based on the time of reporting the preset embedding point.

[0025] A page session (i.e., pageSeq) refers to a continuous, uninterrupted cycle of interaction between a user on a client (such as a browser or app) and the same page instance; that is, the complete interaction process from entering the target page to naturally leaving or exiting the device. When the target page successfully loads and becomes interactive, a new page session begins, and a unique page session identifier is generated.

[0026] In one or more embodiments of this specification, the target page corresponding to the page session is monitored to check for page session reset events such as background reactivation or page refresh. If a page session reset event exists, the page session is terminated, and the next page session is started, generating a page sequence identifier for the next page session. This limitation to obtaining the current behavior trajectory and the complete behavior trajectory within the same page session avoids cross-session risks and ensures that the current behavior trajectory is a prefix of the complete behavior trajectory. This lays a reliable foundation for subsequent page state calibration based on backtracking of the complete behavior trajectory.

[0027] The current behavior trajectory is a sequence of user behavior data within the same page session up to the time of the preset tracking point reporting. For example... Figure 2As shown, during a page session, the user's behavioral trajectory over a period of time (i.e., the user behavioral trajectory at the Session level) includes actions such as clicking, scrolling up and down, long-pressing, and switching left and right. By pre-setting event tracking points, a sequence of user behavior data ordered chronologically up to the time the pre-set event tracking point was reported can be obtained. For example, in a certain application scenario, the current behavioral trajectory of the first type of page session is Click-a->Click-b->Click-d. The current behavioral trajectory of the second type of page session is Click-c->Click-e. The user behavior data package contains user behavior types such as clicks, scrolling, and long-presses, timestamps, spatial coordinates corresponding to the user behavior events, and the target elements of the interacted elements such as buttons and sliders.

[0028] Within a single page session, multiple different current behavior trajectories may be generated due to the reporting times of multiple pre-set event tracking points. The setting of pre-set event tracking point reporting times is flexible and business-oriented, and can be configured based on the business requirements of the current target page. For example, in a certain application scenario, it can be set to trigger periodically based on business requirements, automatically reporting once every 30 seconds, generating a current behavior trajectory from the start of the page session up to that point.

[0029] A complete behavior trajectory is a sequence of user behavior data arranged chronologically within the same page session, marked by the same page sequence identifier, from the beginning to the natural end or middle of the session. For example, in a certain application scenario, the complete behavior trajectory of the first type of page session includes the current behavior trajectory of the first page session as Click-a->Click-b->Click-d. The complete behavior trajectory of the second type of page session includes the current behavior trajectory of the second page session as Click-c->Click-e.

[0030] S104: Based on the complete behavior trajectory of the user terminal, determine the adjacent subsequent behaviors of the current behavior trajectory.

[0031] Since the current behavior trajectory is always a prefix of the complete behavior trajectory, the current page state is roughly calibrated by using the last behavior of the current behavior trajectory and its adjacent subsequent behavior. The adjacent subsequent behavior is the first user behavior data following the current behavior trajectory that meets the validity criteria: it is a user-initiated interaction behavior, and the time interval occurs within the same interaction context within a reasonable window. For example, if the current behavior trajectory includes [A1, A2, A3…An], and the complete behavior trajectory includes [A1, A2, A3…An, B1, B2…Bm], and user behavior data B1 immediately adjacent to the current behavior trajectory is valid, then user behavior data B1 is the adjacent subsequent behavior of the current behavior trajectory.

[0032] In one or more embodiments of this specification, the adjacent subsequent behaviors of the current behavior trajectory are determined based on the complete behavior trajectory of the user terminal, specifically through the following methods: Within the complete behavior trajectory, find the behavior trajectory segment corresponding to the final behavior of the current behavior trajectory. The final behavior is the last action of the current behavior trajectory. For example, if the current behavior trajectory includes [A1, A2, A3...An], then An is the final behavior of the current behavior trajectory. Within the complete behavior trajectory of the dialogue on the same page as the current behavior trajectory, find the position that matches the final behavior An of the current behavior trajectory. This match means that the behavior type of the final behavior An in the complete behavior trajectory is the same, the target element identifier of the behavior is the same, and the timestamp difference is within a reasonable error range. After determining the position that matches the final behavior An, obtain the behavior trajectory segment immediately adjacent to that position based on the time sequence of the complete behavior trajectory.

[0033] A behavior trajectory segment can be a logically continuous interaction fragment located after the final behavior of the current behavior trajectory, divided chronologically. The boundaries of this logically continuous interaction fragment can be defined according to specific business rules. For example, in a certain scenario, it might be defined as all user behavior data until a preset task completion event, such as successful order payment or pop-up confirmation closing, or a page state change, such as page redirection, is detected. By acquiring this behavior trajectory segment, it can be ensured that the final behavior of the current behavior trajectory and the behaviors within this segment belong to the same context. For example, in an e-commerce shopping scenario, if the final behavior of the current behavior trajectory is "clicking the coupon button," then its corresponding behavior trajectory segment is all user behavior data recorded from the moment the click occurs until the coupon pop-up closes or the page redirection event occurs.

[0034] After obtaining the behavior trajectory segment, the user behavior data adjacent to the final behavior in that behavior trajectory segment is extracted as the adjacent subsequent behavior of the current behavior trajectory. In other words, the next user behavior data adjacent to the final behavior in the behavior trajectory segment, i.e., the user-triggered behavior operation data, is obtained based on the timestamp order and is used as the adjacent subsequent behavior.

[0035] Following the example above, in an e-commerce shopping scenario, the determined behavior trajectory segment after clicking to claim a coupon sequentially records "pop-up coupon list pop-up window", "user scrolls through browser list", and "user clicks the 'Use Now' button on one of the coupons". Since the pop-up coupon list pop-up window is not considered user behavior data, the first user behavior data adjacent to the end trajectory of the current behavior trajectory, "user scrolls through browser list", is the adjacent subsequent behavior of the current behavior trajectory.

[0036] S106: Determine the state probability of each page state by the end behavior of the current behavior trajectory and the adjacent subsequent behavior, and calibrate the page state corresponding to the current behavior trajectory.

[0037] As mentioned above, the current method records all user actions from entering the page to the current moment in sequence, forming a behavior sequence. This sequence is then compared with a predefined page state machine describing page state transitions to identify the page's current state. However, this method is problematic in complex interaction scenarios. For example, if the existing simple page state machine is a-s1->b-s2->c-s3->d-s4->e, and the user behavior sequence is s1, s2, s4, then matching the first two user behavior data (s1, s2) with the page state machine determines the user is in page state c. However, the third user behavior data (s4) indicates a transition from page state d to page state e, so the current page state is directly determined to be d. Therefore, the valid outgoing edge s4 of page state c inferred from the first two user behavior data does not match the next input user behavior data (s4), making it difficult to explain the execution of s4 from page state c. Furthermore, in complex business scenarios, this problem also presents the issue of abnormal branches during state transitions causing judgment errors. For example, in one embodiment, the behavior sequence Seq1 is Click-a->Click-b, according to... Figure 3 The page state machine shown indicates that there are multiple possibilities for the state transition based on Seq1, with probabilities falling between "Page State 1.1", "Page State 1.2", and "Page State 1.3". In a specific business scenario, the transition from "Page State 1.1" to "Page State 1.2" can be considered a failed state transition due to an interface exception; the transition from "Page State 1.1" to "Page State 1.2" is the expected state transition; and the transition from "Page State 1.1" to "Page State 1.3" can be a blank screen or a fallback page caused by a JS exception. If the failed state transition due to an exception or the transition to an unexpected state is not identified, unexplained hotspots will appear in the heatmap view for a certain state, thus interfering with the data analysis of the heatmap. Secondly, there is also the problem of path tracing failure due to the uncertainty of the page initialization state, such as when the behavior sequence Seq2 is Click-c. Similarly, based on... Figure 3It can be seen that there are multiple possibilities for the state based on Seq2 transition, with a probability between "page state 2.1" and "page state 3.1". In a specific business scenario, it can be assumed that: during page initialization, a pop-up window (page state 2.0) or a task panel (page state 3.0) may appear directly, and both the pop-up window and the task panel have the same action point (such as opening the asset panel to view rewards). After clicking, the prize list is viewed under different tabs in the asset panel (page state 2.1, page state 3.1). If the state at which the page is initialized is not identified, the state after the switch cannot be determined.

[0038] Based on this, in one or more embodiments of this description, to solve the heatmap error problem caused by interface anomalies, JS anomalies, or uncertain initialization states in complex state machine scenarios, after obtaining the current behavior trajectory and the adjacent subsequent behaviors corresponding to the current behavior trajectory in the complete behavior trajectory based on the above steps S102 and S104, since the current behavior trajectory must be a prefix trajectory of the complete behavior trajectory, the page state can be inferred by the terminal behavior (i.e., the last behavior) and the adjacent subsequent behaviors corresponding to the current behavior trajectory. Furthermore, by combining the probability statistics of the page state, a high-probability state is selected to calibrate the page state corresponding to the current behavior trajectory. Specifically, combined with... Figure 4 As shown, the state probability of each page state is determined by the terminal behavior of the current behavior trajectory and the adjacent subsequent behaviors, and the page state corresponding to the current behavior trajectory is calibrated, including the following steps: S402: Based on the terminal behavior and the adjacent subsequent behavior, form a behavior pair for the current behavior trajectory; S404: In the preset page state machine, match the state transition path corresponding to the behavior pair; S406: Determine the state probability of each state transition path based on the occurrence probability of each state transition path and the conditional probability of the action pair; S408: Based on the state probability of the state transition path, calibrate the page state corresponding to the current behavior trajectory.

[0039] Based on the problems described above, it is clear that the traditional method of combining the current behavior trajectory with the page state machine only matches the end behavior of the current behavior trajectory, resulting in state ambiguity. That is, the same operation may be triggered from multiple different states. Knowing only the end behavior, it is impossible to determine whether the end behavior corresponds to a normal flow state or an abnormal state. Furthermore, since it is unknown from which initialization state the end behavior started, it is impossible to determine which page state the end behavior will trigger during normal flow. To address this, when the end behavior of the current behavior trajectory is labeled as behavior A, and the adjacent subsequent behavior of this end behavior from the complete behavior trajectory is identified as behavior B, these two are grouped into an ordered binary tuple as the behavior pair of the current behavior trajectory. For example, if the end behavior is Click-c and the adjacent subsequent behavior is Click-e, then the behavior pair (Click-c, Click-e) is formed.

[0040] A state transition path is a directed path consisting of page state transitions that matches the behavior pair of the current behavior trajectory within a pre-defined page state machine. In the pre-defined page state machine, nodes represent page states, and edges represent transitions between states. The behavior pair of the current behavior trajectory obtained in the previous step is used as a query condition and searched within the pre-defined page state machine to match the state transition path corresponding to that behavior pair. Specifically, within the pre-defined page state machine, the first transition edge is found to match the final behavior, and a second transition edge exists in the page state of the first transition edge, with the triggering condition of the second transition edge matching the adjacent subsequent behavior in the behavior pair. For each matched state transition path, the prior probability of its corresponding starting page state as the context of the behavior occurrence in all historical sessions is obtained. For example, based on historical statistics, the frequency of "page state 2.1" as the starting point is determined to be 60%, and "page state 3.1" to be 40%. Simultaneously, for each path, the conditional probability of the observed behavior occurring in the initial state described by that path is calculated. For example, the probability of this behavior occurring in "page state 2.1" is 90%, and the probability of this behavior occurring in "page state 3.1" is 70%. Then, the occurrence probability and conditional probability of each state transition path are combined to calculate the state probability of that state transition path. Continuing with the example above, the probability of "page state 2.1" is 66%, and the probability of "page state 3.1" is 34%.

[0041] Then, based on the state probability, the page state after calibration of the current behavior trajectory is selected. This involves obtaining the starting page state corresponding to each state transition path, where the starting page state is the page state corresponding to the beginning of a state transition path, i.e., the page state starting from the end behavior of the current behavior trajectory. Therefore, based on the state probability of the state transition paths, it is determined whether the current behavior trajectory is an abnormal trajectory. This is done by comparing the state probability of each state transition path with an anomaly judgment threshold. If the state probability of a state transition path is lower than the anomaly judgment threshold, the current behavior trajectory is determined to be abnormal and marked as an abnormal trajectory.

[0042] The anomaly detection threshold is a high-confidence threshold set by data analysts, or it can be set through other reliable methods, such as a high-confidence threshold determined based on historical data. For example, the calibration process is performed on a historical dataset to obtain the state probability of the behavioral trajectory, and a curve showing the relationship between the state probability and the calibration accuracy is plotted. The anomaly detection threshold is set on this curve to correspond to the probability value when the calibration accuracy first reaches or exceeds a preset target (determined based on business needs).

[0043] If this is not an abnormal trajectory, then based on the state probabilities obtained above, and based on the high-probability page states, the starting page state with the highest state probability is used as the page state after the current behavior trajectory is calibrated. For example... Figure 5 The pre-defined page state machine in a certain application scenario is shown, which combines the "page state 1.4" corresponding to the adjacent subsequent action Click-d, as shown in the example. Figure 6 The backtracking shown indicates that the current behavior trajectory's page state is "Page State 1.2". This can be further confirmed by combining this with the adjacent subsequent behavior, Click-e, which corresponds to "Page State 2.2". Figure 6 The backtracking shown indicates that the current behavior trajectory's page state is "Page State 2.1".

[0044] Furthermore, it should be noted that methods based on complete behavioral trajectories cannot perfectly calibrate random errors during state transitions, such as... Figure 7 The arrival actions for "Page State 2.1" and "Page State 3.1" are the same (Click-c), and they can transition to "Page State 2.2" through the same action (Click-e). Even by tracing the complete action path back (Click-c -> Click-e), the current page state cannot be determined. However, in actual business scenarios, the probability of this path being triggered is low. Therefore, the method also includes: if the current action path is an abnormal path, then mark the current action path as invalid data, thereby reducing the workload and computational complexity of data statistics and ensuring the accuracy of the heatmap.

[0045] This process uses the action pairs formed by the final action of the current behavior trajectory and adjacent subsequent actions to match the subsequent actions against the preset page state machine. This overcomes the problem of state ambiguity when relying solely on the final action of the current behavior trajectory for state matching. By expanding the contextual information of the current behavior trajectory through adjacent subsequent actions, the accuracy of state judgment is improved by reverse derivation and backtracking of the previous state. By identifying and marking abnormal trajectories, these trajectories can be directly discarded as invalid data, thereby reducing noise interference in the statistical data of user behavior and improving the accuracy of analysis results such as heatmaps.

[0046] Furthermore, in another embodiment, after identifying abnormal trajectories, in addition to directly filtering out the current behavior trajectories corresponding to the abnormal trajectories, in order to maximize the use of user behavior data whose page states cannot be calibrated due to complexity and anomalies while maintaining overall data reliability, and to avoid the problem of valuable uncertain information being filtered out, the method also includes, as... Figure 8 The following steps are shown: S802: If the current behavior trajectory is an abnormal trajectory, then compare the state probability of the abnormal trajectory with a preset low confidence threshold; S804: Mark the current behavior trajectory where all state probabilities are less than the preset low confidence threshold, or where the state transition path is empty, as invalid data; S806: Obtain the fuzzy behavior trajectory in the abnormal trajectory excluding the invalid data, and determine the weight of each starting page state based on the state probability of each state transition path corresponding to the fuzzy behavior trajectory. S808: Based on the weights of each of the initial page states, obtain the composite page state after the current behavior trajectory calibration.

[0047] If the current behavior trajectory is an abnormal trajectory, the state probability of the abnormal trajectory is compared with a preset low-confidence threshold. If the state probability of each starting page state corresponding to the behavior pair of the current behavior trajectory is less than the preset low-confidence threshold, it indicates that the current behavior trajectory does not match any of the known normal states of the preset page state machine, and there is a serious anomaly. Furthermore, if the state transition path of the behavior pair of the current behavior trajectory in the preset page state machine is empty, that is, no starting page state is matched, it indicates that the current behavior trajectory has an inexplicable anomaly. Therefore, these two types of abnormal trajectories are marked as invalid data and will be filtered out when generating heatmaps in the future, thereby avoiding the problem of abnormal data contaminating the heatmap video.

[0048] A preset low-confidence threshold is used to identify invalid data or serious anomalies. This threshold is a low value set by data analysts or based on current business quality. Additionally, a page state calibration process can be performed on historical datasets to obtain the state transition paths and probabilities matched to each current behavior trajectory. Then, by analyzing error logs, customer feedback, or manually annotated abnormal trajectories, and comparing them with trajectories that cannot be matched with state transition paths in the preset page state machine, the corresponding state probability distribution is determined. The state probability corresponding to the appropriate state probability distribution value is used as the preset low-confidence threshold based on the current business scenario requirements. For example, if a 95% state probability distribution value is selected based on business scenario requirements, and the best state probability obtained after calibration is lower than this threshold, then there is a 95% probability that it belongs to abnormal or invalid data that should be discarded. Using the state probability corresponding to this distribution value as the preset low-confidence threshold can maximize the elimination of noise to ensure the accuracy of the heatmap while also cautiously recovering some information value from uncertain data, achieving a better balance between data utilization efficiency and result reliability.

[0049] After filtering out invalid data, the remaining fuzzy behavioral trajectories marked as abnormal trajectories, excluding invalid data, are obtained from the abnormal trajectories. It can be understood that a fuzzy behavioral trajectory is the current behavioral trajectory among the abnormal trajectories excluding invalid data. Although the fuzzy behavioral trajectory exceeds the aforementioned anomaly judgment threshold, meaning it does not meet the high confidence standard, its state probability is not so low as to be directly discarded, and at least one matching path can be found in the state machine; it's just that the matched path is not unique or the confidence is insufficient.

[0050] The state probabilities of each state transition path corresponding to the fuzzy behavior trajectory are obtained. The method for obtaining these state probabilities is consistent with the method described in S406 above. After obtaining the state probabilities, the normalized probability value of each state probability is used as the weight of the corresponding starting page state. For example, if a fuzzy behavior trajectory has two state transition paths, with a probability of 0.4 for starting page state s1 and 0.4 for starting page state s2, then the normalized weight of starting page state s1 is 0.5, and the weight of starting page state s2 is 0.5. Based on the determined weights of each starting page state, each starting page state and its corresponding weight are encapsulated to obtain the composite page state after calibration of the current behavior trajectory.

[0051] S108: Generate a page heatmap based on the calibrated page status and the user behavior data.

[0052] Existing heatmap generation solutions typically aggregate user behavior data based directly on uncalibrated page states or through simple matching of current behavior trajectories with page states obtained from state page states. However, in complex interaction scenarios, this approach can lead to distortion in the generated heatmap due to errors in page states. For example, when page states are uncertain, user behavior data belonging to different page states may be incorrectly grouped into the same page state's heatmap layer. For instance, the heatmap of repeatedly clicking "Retry" under "Page State 1.1" due to an interface anomaly may be superimposed on the heatmap of clicking "Confirm" under "Page State 1.2" in a normal process, resulting in an inexplicable and distorted heatmap value that deviates from the true user intent.

[0053] Therefore, after steps S102-S106 above, through complete behavioral trajectory backtracking and probability statistics-assisted calibration, to solve the heatmap error problem caused by interface anomalies, JS anomalies, or uncertain initialization states in complex state machine scenarios, in one or more embodiments of this specification, based on the calibrated page state and assuming the page state is determined, a page heatmap is generated based on user behavior data. Specifically, generating a page heatmap based on the calibrated page state and the user behavior data includes the following steps: Based on the calibrated page states corresponding to each user behavior data point, the user behavior data is divided into various page state levels. These page state levels include at least: a main interface level corresponding to the basic page state without modal component coverage; a modal interaction level corresponding to the open state of forced-focus components such as pop-ups, drawers, and alerts; and a fuzzy level corresponding to data with calibration results of low-confidence probability labels or anomaly markers. By dividing the user behavior data into these page state levels, a foundation is laid for generating a clearly defined heatmap.

[0054] To ensure that thermal calculations within each level are performed in their corresponding visual context, the coordinates of the corresponding display window—that is, the actual display area on the screen—are obtained for each page state level after division. For example, for the main interface level, the display window is typically the entire browser or app view window. For the modal interaction level, the display window is the visual boundary of the modal component, such as a pop-up or drawer. For ambiguous or abnormal levels, approximate confirmation can be made based on the main interface window or the area where the original coordinates are located. This converts the global page coordinates of user behavior data in each page state level into local coordinates in the corresponding display window coordinate system. This coordinate conversion process can be obtained based on the global page coordinates and the diagonal coordinates of the window in each display window coordinate system; for example, by subtracting the top-left corner coordinate in the display window coordinate system from the global page coordinates, the local coordinates are obtained.

[0055] Based on the local coordinates corresponding to the user behavior data, the spatial distribution density of user behavior interaction points within the display window of each page state level is obtained, thereby obtaining a heatmap density matrix for each page state level. This heatmap density matrix is ​​then color-rendered to generate a heatmap layer corresponding to each page state level. Finally, based on the heatmap layers corresponding to each page state level, a heatmap is generated as follows: Figure 9 The page heatmap shown.

[0056] Furthermore, the process of generating a page heatmap based on the heatmap layers corresponding to each page state level includes the following steps: Based on the stacking style data corresponding to each page state level, the heatmap layers are sorted and rendered sequentially to obtain a page heatmap that restores the visual occlusion relationship. Generally, the main interface layer is at the bottom, and the modal interaction layer is at the top according to its opening order and z-index value. The heatmap layers are rendered onto the final canvas in a bottom-to-top order. This rendering of each heatmap layer to obtain the page heatmap includes: Determine the interaction time interval for each page state level within the page session. The interaction time interval of a page session is the continuous time interval during which a page state level, such as a "coupon pop-up," is at the forefront of the screen and receives user input during a complete page visit. It can be determined by using recorded user behavior timestamps and page state transition events to determine the start and end times of each page state level's position as the screen's interactive focus on the page session's timeline. For example, when no modal is open, the main interface layer is in an interactive state; once a modal is opened, the main interface layer ends its interaction. This start-to-end time period is the interaction time interval for that page state level.

[0057] In real-world scenarios, when a pop-up window opens, the user's attention is primarily focused on the pop-up. Therefore, when rendering a heatmap, not all heatmap layers are rendered uniformly. To differentiate between heatmap layers based on the user's actual attention at a given time, this embodiment determines the attention attenuation coefficient for each heatmap layer based on the relationship between the interaction time interval and the current rendering time. Specifically, it assigns a corresponding attention attenuation coefficient to each heatmap layer based on whether the interaction time interval falls within the current rendering time, for example, by determining the assigned attenuation coefficient based on the proportion of the interaction time interval within the current rendering time. Using these assigned attenuation coefficients, the visual output values ​​of each heatmap layer are weighted and then superimposed to obtain the page heatmap during sequential rendering. This ensures that the heatmap of the layer the user is currently focusing on is highlighted, while the heatmap of non-focused or historically unfocused layers is appropriately faded or hidden, accurately and intuitively reflecting the user's interactive heatmap data in complex modal interfaces.

[0058] Specifically, following Figure 8 In one embodiment, a page heatmap is generated based on the calibrated page state and the user behavior data, specifically including: Based on the calibrated composite page states, the heatmap values ​​of user behavior data are split according to weighted proportions and assigned to the corresponding state heatmap matrices to obtain the state heatmap matrices for each starting page state. For example, if the current behavior trajectory in a certain scenario is calibrated to a composite page state (starting page state s1, weight 0.7; starting page state s2, weight 0.3), then the heatmap value of this user behavior data is split into 0.7 units of heatmap value, which are accumulated in the state heatmap matrix corresponding to state s1, and 0.3 units of heatmap value are accumulated in the state heatmap matrix corresponding to state s2.

[0059] To differentiate the heatmap representations of different functional areas and ensure that heatmap values ​​from different states are correctly rendered to their respective physical regions during heatmap synthesis, the target page is segmented based on its Document Object Model (DOM) to obtain multi-page regions. The DOM is a tree-like data structure built by the browser in memory to represent the page structure and content. The DOM of the target page represents the DOM structure of the specific page instances the user interacts with during the current page session. This process lays the foundation for subsequently associating heatmaps with specific page locations and generating accurate hierarchical heatmaps.

[0060] The process involves obtaining the association between the initial page state and page regions. Since the same physical screen location corresponds to different interactive elements and user intents in different page states, to improve the accuracy of heatmap rendering, this association is achieved by performing similarity matching between the state semantic features corresponding to each initial page state and the functional semantic features corresponding to each page region. This determines which page regions the initial page state primarily associates with or influences. This association can be a one-to-one, one-to-many, or many-to-one mapping, clearly identifying the page regions affected by each initial page state.

[0061] Based on this correlation and correlation degree, the heat values ​​recorded in the heatmap of each state are assigned to the coordinates of the corresponding page area to obtain the heat density matrix corresponding to the starting page state. This heat density matrix reflects the distribution information of the heat values ​​of user behavior data in the two-dimensional space of the page under each starting page state. Then, this heat density matrix is ​​rendered to generate a page heatmap. This page heatmap distinguishes and displays the heat distribution under different starting page states.

[0062] Furthermore, in one or more embodiments of this specification, the method further includes: Based on page heatmaps and preset heatmap thresholds, dense and sparse user interaction areas are identified under different page states. Dense user interaction areas are continuous regions in the page heatmap corresponding to a given initial page state where the heat value exceeds the preset first heatmap threshold, representing hotspots frequently clicked, swiped, or followed by users in that state. Sparse user interaction areas are continuous regions in the page heatmap of the same state where the heat value is below the preset second heatmap threshold, representing areas with less user interaction in that state. The preset first and second heatmap thresholds are obtained statistically or determined based on business objectives. To provide suggestions for subsequent page optimization based on the difference between actual and expected user hotspots, this embodiment determines the difference characteristics between dense and sparse user interaction areas based on the correlation between the initial page state and page regions. These difference characteristics include the difference between theoretical and actual correlation strength. In order to determine whether densely populated and sparsely populated user interaction areas on a page conform to the page design intent and to ascertain the degree of deviation, thus providing direction for subsequent optimization, the correlation between the initial page state and page areas was obtained. This is the theoretical correlation degree between each page area in each initial page state. Based on this correlation degree, it was compared with the current densely populated and sparsely populated user interaction areas to determine the difference between the theoretical and actual correlation strength. Based on the optimization direction matched by this difference feature, corresponding optimization suggestions were generated to facilitate subsequent page optimization adjustments.

[0063] Based on the same idea, one or more embodiments of this specification also provide apparatus and devices corresponding to the above methods, such as... Figure 10 , Figure 11 As shown.

[0064] Figure 10 This specification provides a schematic diagram of a thermal data calibration device according to one or more embodiments, the device comprising: The acquisition module 1002 is used to acquire the current behavior trajectory and the complete behavior trajectory of a session on the same page; the current behavior trajectory is a sequence of user behavior data based on the time of reporting the preset embedding point; The determination module 1004 is used to determine the adjacent subsequent behaviors of the current behavior trajectory based on the complete behavior trajectory of the user terminal; The calibration module 1006 is used to determine the state probability of each page state by the end behavior of the current behavior trajectory and the adjacent subsequent behavior, and to calibrate the page state corresponding to the current behavior trajectory. The generation module 1008 is used to generate a page heatmap based on the calibrated page status and the user behavior data.

[0065] Optionally, the calibration module 1006 specifically includes: Based on the terminal behavior and the adjacent subsequent behavior, a behavior pair is formed to constitute the current behavior trajectory; In the preset page state machine, match the state transition path corresponding to the behavior pair; The state probability of each state transition path is determined based on the occurrence probability of each state transition path and the conditional probability of the behavior pair. Based on the state probability of the state transition path, the page state corresponding to the current behavior trajectory is calibrated.

[0066] Optionally, the calibration module 1006 specifically includes: Obtain the starting page state corresponding to each of the aforementioned state transition paths; Based on the state probability of the state transition path, determine whether the current behavior trajectory is an abnormal trajectory; If it is not an abnormal trajectory, then based on the state probability, the corresponding starting page state is selected as the page state after the current behavior trajectory is calibrated.

[0067] Optionally, the device further includes The marking module 1010 is used to mark the current behavior trajectory as invalid data if the current behavior trajectory is an abnormal trajectory.

[0068] Optionally, the device further includes: a composite page state calibration module 1012; The composite page state calibration module 1012 is used to compare the state probability of the abnormal trajectory with a preset low confidence threshold if the current behavior trajectory is an abnormal trajectory. The current behavior trajectory whose state probabilities are all less than the preset low confidence threshold, or whose state transition path is empty, is marked as invalid data; Obtain the fuzzy behavior trajectory in the abnormal trajectory excluding the invalid data, and determine the weight of each starting page state based on the state probability of each state transition path corresponding to the fuzzy behavior trajectory; Based on the weights of each of the initial page states, the composite page state after the current behavior trajectory calibration is obtained.

[0069] Optionally, the device further includes: The listening module 1014 is used to uniquely identify the page session based on the page sequence identifier; Monitor the target page corresponding to the page session to see if a page session reset event exists; the page session reset event includes: background reactivation and page refresh; If it exists, the next page session is started, and a page sequence identifier for the next page session is generated.

[0070] Optionally, the generation module 1008 specifically includes: Based on the calibrated page state corresponding to each user behavior data, the user behavior data is divided into each page state level; Based on the display window coordinates corresponding to each page state level, the coordinates of the user behavior data are transformed to obtain local coordinates; Based on the local coordinates corresponding to the user behavior data, a heat density matrix for each page state level is obtained to generate a heat layer corresponding to each page state level. A page heatmap is generated based on the heatmap layers corresponding to each page state level.

[0071] Optionally, the generation module 1008 specifically includes: Based on the stacking style data corresponding to each page state level, the heatmap layers are sorted and rendered sequentially to obtain a page heatmap. The process of rendering each heatmap layer to obtain a page heatmap includes: Determine the interaction time interval of each page state level in the page session; Based on the relationship between the interaction time interval and the current rendering time, the attention decay coefficient of each heat layer is determined; The heatmap of the page is obtained by weighting the corresponding heatmap layers using the attention attenuation coefficient and then overlaying them for rendering.

[0072] Optionally, the acquisition module 1002 specifically includes: Based on the preset event points, a sequence of user behavior data up to the reporting time of each preset event point is obtained, which constitutes the current behavior trajectory. A complete behavioral trajectory is constructed based on the user behavior data sequence corresponding to the page session.

[0073] Optionally, the determining module 1004 specifically includes: In the complete behavior trajectory, find the behavior trajectory segment corresponding to the end behavior of the current behavior trajectory; Extract user behavior data adjacent to the terminal behavior in the behavior trajectory segment, and use it as the adjacent subsequent behavior of the current behavior trajectory.

[0074] Optionally, the calibration module 1006 specifically includes: Based on the calibrated composite page state, the heat values ​​corresponding to the user behavior data are allocated to obtain the state heat matrix of each of the initial page states; Based on the document object model of the target page corresponding to the page session, the target page is segmented to obtain multi-page regions; Obtain the association between the starting page state and the page area; Based on the aforementioned relationship, each state thermal matrix is ​​assigned to the coordinates of the page area to obtain the thermal density matrix corresponding to the initial page state. The thermal density matrix is ​​rendered to generate a page heatmap; the page heatmap distinguishes and displays the thermal distribution under different initial page states.

[0075] Optionally, the device further includes: an optimization module 1016; The optimization module 1016 is used to identify dense user interaction areas and sparse user interaction areas under different page states based on the page heatmap and the preset heat threshold. Based on the association between the starting page state and the page area, determine the difference characteristics between the dense user interaction area and the sparse user interaction area; the difference characteristics include the difference between the theoretical association strength and the actual association strength; Based on the aforementioned differences, optimization suggestions are generated.

[0076] Figure 11 A schematic diagram of a thermal data calibration device provided for one or more embodiments of this specification, the device comprising: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enable the at least one processor to: Obtain the current behavior trajectory and complete behavior trajectory of the session on the same page; the current behavior trajectory is a sequence of user behavior data up to the time of the preset data entry point reporting. Based on the complete behavioral trajectory of the user terminal, determine the adjacent subsequent behaviors of the current behavioral trajectory; By combining the final behavior of the current behavior trajectory with the adjacent subsequent behavior, the state probability of each page state is determined, and the page state corresponding to the current behavior trajectory is calibrated. A page heatmap is generated based on the calibrated page status and the user behavior data.

[0077] In the 1990s, improvements to a technology could be clearly distinguished as either hardware improvements (e.g., improvements to the circuit structure of diodes, transistors, switches, etc.) or software improvements (improvements to the methodology). However, with technological advancements, many methodological improvements today can be considered direct improvements to the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved methodology into the hardware circuit. Therefore, it cannot be said that a methodological improvement cannot be implemented using hardware physical modules. For example, a Programmable Logic Device (PLD) (such as a Field Programmable Gate Array (FPGA)) is such an integrated circuit whose logic function is determined by the user programming the device. Designers can program and "integrate" a digital system onto a PLD themselves, without needing chip manufacturers to design and manufacture dedicated integrated circuit chips. Furthermore, nowadays, instead of manually manufacturing integrated circuit chips, this programming is mostly implemented using "logic compiler" software. Similar to the software compiler used in program development, the original code before compilation must also be written in a specific programming language, called a Hardware Description Language (HDL). There are many HDLs, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, and RHDL (Ruby Hardware Description Language). Currently, the most commonly used are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also understand that by simply performing some logic programming on the method flow using one of these hardware description languages ​​and programming it into an integrated circuit, the hardware circuit implementing the logical method flow can be easily obtained.

[0078] The controller can be implemented in any suitable manner. For example, it can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicon Labs C8051F320. A memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also recognize that, in addition to implementing the controller in purely computer-readable program code form, the same functionality can be achieved by logically programming the method steps to make the controller take the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, such a controller can be considered a hardware component, and the means included therein for implementing various functions can also be considered as structures within the hardware component. Alternatively, the means for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0079] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0080] For ease of description, the above devices are described in terms of function, divided into various units. Of course, in implementing this specification, the functions of each unit can be implemented in one or more software and / or hardware components.

[0081] Those skilled in the art will understand that the embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, the embodiments of this specification can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the embodiments of this specification can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0082] This specification is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this specification. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0083] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0084] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0085] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0086] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0087] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0088] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0089] This specification can be described in the general context of computer-executable instructions that are executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This specification can also be practiced in distributed computing environments, where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0090] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, devices, and non-volatile computer storage media are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0091] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0092] The above description is merely one or more embodiments of this specification and is not intended to limit this specification. Various modifications and variations can be made to the one or more embodiments of this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of one or more embodiments of this specification should be included within the scope of the claims of this specification.

Claims

1. A method for calibrating a heatmap, the method comprising: Get the current behavior trajectory and complete behavior trajectory of a session on the same page; The current behavior trajectory is a sequence of user behavior data up to the time of reporting the preset embedded point; Based on the complete behavioral trajectory of the user terminal, determine the adjacent subsequent behaviors of the current behavioral trajectory; By combining the final behavior of the current behavior trajectory with the adjacent subsequent behavior, the state probability of each page state is determined, and the page state corresponding to the current behavior trajectory is calibrated. A page heatmap is generated based on the calibrated page status and the user behavior data.

2. The method as described in claim 1, wherein the state probability of each page state is determined by the terminal behavior of the current behavior trajectory and the adjacent subsequent behavior, and the page state corresponding to the current behavior trajectory is calibrated, specifically including: Based on the terminal behavior and the adjacent subsequent behavior, a behavior pair is formed to constitute the current behavior trajectory; In the preset page state machine, match the state transition path corresponding to the behavior pair; The state probability of each state transition path is determined based on the occurrence probability of each state transition path and the conditional probability of the behavior pair. Based on the state probability of the state transition path, the page state corresponding to the current behavior trajectory is calibrated.

3. The method as described in claim 2, wherein calibrating the page state corresponding to the current behavior trajectory based on the state probability of the state transition path, specifically includes: Obtain the starting page state corresponding to each of the aforementioned state transition paths; Based on the state probability of the state transition path, determine whether the current behavior trajectory is an abnormal trajectory; If it is not an abnormal trajectory, then based on the state probability, the corresponding starting page state is selected as the page state after the current behavior trajectory is calibrated.

4. The method of claim 3, further comprising: If the current behavior trajectory is an abnormal trajectory, then the current behavior trajectory will be marked as invalid data.

5. The method of claim 3, further comprising: If the current behavior trajectory is an abnormal trajectory, then compare the state probability of the abnormal trajectory with a preset low confidence threshold; The current behavior trajectory whose state probabilities are all less than the preset low confidence threshold, or whose state transition path is empty, is marked as invalid data; Obtain the fuzzy behavior trajectory in the abnormal trajectory excluding the invalid data, and determine the weight of each starting page state based on the state probability of each state transition path corresponding to the fuzzy behavior trajectory; Based on the weights of each of the initial page states, the composite page state after the current behavior trajectory calibration is obtained.

6. The method of claim 1, further comprising: The page session is uniquely identified based on the page sequence identifier; Monitor the target page corresponding to the page session to see if a page session reset event exists; The page session reset events include: background reactivation and page refresh; If it exists, the next page session is started, and a page sequence identifier for the next page session is generated.

7. The method as described in claim 1, wherein generating a page heatmap based on the calibrated page state and the user behavior data specifically includes: Based on the calibrated page state corresponding to each user behavior data, the user behavior data is divided into each page state level; Based on the display window coordinates corresponding to each page state level, the coordinates of the user behavior data are transformed to obtain local coordinates; Based on the local coordinates corresponding to the user behavior data, a heat density matrix for each page state level is obtained to generate a heat layer corresponding to each page state level. A page heatmap is generated based on the heatmap layers corresponding to each page state level.

8. The method as described in claim 7, wherein generating a page heatmap based on the heatmap layers corresponding to each page state level specifically includes: Based on the stacking style data corresponding to each page state level, the heatmap layers are sorted and rendered sequentially to obtain a page heatmap. The process of rendering each heatmap layer to obtain a page heatmap includes: Determine the interaction time interval of each page state level in the page session; Based on the relationship between the interaction time interval and the current rendering time, the attention decay coefficient of each heat layer is determined; The heatmap of the page is obtained by weighting the corresponding heatmap layers using the attention attenuation coefficient and then overlaying them for rendering.

9. The method as described in claim 1, obtaining the current behavior trajectory and complete behavior trajectory of a user's session on the same page, specifically includes: Based on the preset event points, a sequence of user behavior data up to the reporting time of each preset event point is obtained, which constitutes the current behavior trajectory. A complete behavioral trajectory is constructed based on the user behavior data sequence corresponding to the page session.

10. The method as described in claim 1, wherein determining the adjacent subsequent behaviors of the current behavior trajectory based on the complete behavior trajectory of the user terminal, specifically includes: In the complete behavior trajectory, find the behavior trajectory segment corresponding to the end behavior of the current behavior trajectory; Extract user behavior data adjacent to the terminal behavior in the behavior trajectory segment, and use it as the adjacent subsequent behavior of the current behavior trajectory.

11. The method as described in claim 5, wherein generating a page heatmap based on the calibrated page state and the user behavior data specifically includes: Based on the calibrated composite page state, the heat values ​​corresponding to the user behavior data are allocated to obtain the state heat matrix of each of the initial page states; Based on the document object model of the target page corresponding to the page session, the target page is segmented to obtain multi-page regions; Obtain the association between the starting page state and the page area; Based on the aforementioned relationship, each state thermal matrix is ​​assigned to the coordinates of the page area to obtain the thermal density matrix corresponding to the initial page state. The thermal density matrix is ​​rendered to generate a page heatmap; the page heatmap distinguishes and displays the thermal distribution under different initial page states.

12. The method of claim 11, further comprising: Based on the page heatmap and the preset heat threshold, identify dense user interaction areas and sparse user interaction areas under different page states; Based on the association between the starting page state and the page area, determine the difference characteristics between the dense user interaction area and the sparse user interaction area; the difference characteristics include the difference between the theoretical association strength and the actual association strength; Based on the aforementioned differences, optimization suggestions are generated.

13. A calibration apparatus for a heat map, the apparatus comprising: The acquisition module is used to acquire the current behavior trajectory and the complete behavior trajectory of a session located on the same page; The current behavior trajectory is a sequence of user behavior data up to the time of reporting the preset embedded point; The determination module is used to determine the adjacent subsequent behaviors of the current behavior trajectory based on the complete behavior trajectory of the user terminal; The calibration module is used to determine the state probability of each page state by the end behavior of the current behavior trajectory and the adjacent subsequent behavior, and to calibrate the page state corresponding to the current behavior trajectory. The generation module is used to generate a page heatmap based on the calibrated page status and the user behavior data.

14. The apparatus of claim 13, wherein the calibration module specifically comprises: Based on the terminal behavior and the adjacent subsequent behavior, a behavior pair is formed to constitute the current behavior trajectory; In the preset page state machine, match the state transition path corresponding to the behavior pair; The state probability of each state transition path is determined based on the occurrence probability of each state transition path and the conditional probability of the behavior pair. Based on the state probability of the state transition path, the page state corresponding to the current behavior trajectory is calibrated.

15. The apparatus of claim 14, wherein the calibration module specifically comprises: Obtain the starting page state corresponding to each of the aforementioned state transition paths; Based on the state probability of the state transition path, determine whether the current behavior trajectory is an abnormal trajectory; If it is not an abnormal trajectory, then based on the state probability, the corresponding starting page state is selected as the page state after the current behavior trajectory is calibrated.

16. The apparatus of claim 15, further comprising: The marking module is used to mark the current behavior trajectory as invalid data if the current behavior trajectory is an abnormal trajectory.

17. The apparatus of claim 15, further comprising: Composite page status calibration module; The composite page state calibration module is used to compare the state probability of the abnormal trajectory with a preset low confidence threshold if the current behavior trajectory is an abnormal trajectory. The current behavior trajectory whose state probabilities are all less than the preset low confidence threshold, or whose state transition path is empty, is marked as invalid data; Obtain the fuzzy behavior trajectory in the abnormal trajectory excluding the invalid data, and determine the weight of each starting page state based on the state probability of each state transition path corresponding to the fuzzy behavior trajectory; Based on the weights of each of the initial page states, the composite page state after the current behavior trajectory calibration is obtained.

18. The apparatus of claim 13, further comprising: A listening module is used to uniquely identify the page session based on a page sequence identifier; Monitor the target page corresponding to the page session to see if a page session reset event exists; The page session reset events include: background reactivation and page refresh; If it exists, the next page session is started, and a page sequence identifier for the next page session is generated.

19. The apparatus of claim 13, wherein the generation module specifically comprises: Based on the calibrated page state corresponding to each user behavior data, the user behavior data is divided into each page state level; Based on the display window coordinates corresponding to each page state level, the coordinates of the user behavior data are transformed to obtain local coordinates; Based on the local coordinates corresponding to the user behavior data, a heat density matrix for each page state level is obtained to generate a heat layer corresponding to each page state level. A page heatmap is generated based on the heatmap layers corresponding to each page state level.

20. The apparatus of claim 19, wherein the generation module specifically comprises: Based on the stacking style data corresponding to each page state level, the heatmap layers are sorted and rendered sequentially to obtain a page heatmap. The process of rendering each heatmap layer to obtain a page heatmap includes: Determine the interaction time interval of each page state level in the page session; Based on the relationship between the interaction time interval and the current rendering time, the attention decay coefficient of each heat layer is determined; The heatmap of the page is obtained by weighting the corresponding heatmap layers using the attention attenuation coefficient and then overlaying them for rendering.

21. The apparatus of claim 13, wherein the acquisition module specifically comprises: Based on the preset event points, a sequence of user behavior data up to the reporting time of each preset event point is obtained, which constitutes the current behavior trajectory. A complete behavioral trajectory is constructed based on the user behavior data sequence corresponding to the page session.

22. The apparatus of claim 13, wherein the determining module specifically comprises: In the complete behavior trajectory, find the behavior trajectory segment corresponding to the end behavior of the current behavior trajectory; Extract user behavior data adjacent to the terminal behavior in the behavior trajectory segment, and use it as the adjacent subsequent behavior of the current behavior trajectory.

23. The apparatus of claim 17, wherein the calibration module specifically comprises: Based on the calibrated composite page state, the heat values ​​corresponding to the user behavior data are allocated to obtain the state heat matrix of each of the initial page states; Based on the document object model of the target page corresponding to the page session, the target page is segmented to obtain multi-page regions; Obtain the association between the starting page state and the page area; Based on the aforementioned relationship, each state thermal matrix is ​​assigned to the coordinates of the page area to obtain the thermal density matrix corresponding to the initial page state. The thermal density matrix is ​​rendered to generate a page heatmap; the page heatmap distinguishes and displays the thermal distribution under different initial page states.

24. The apparatus of claim 23, further comprising: Optimize modules; The optimization module is used to identify dense user interaction areas and sparse user interaction areas under different page states based on the page heatmap and preset heat thresholds. Based on the association between the starting page state and the page area, determine the difference characteristics between the dense user interaction area and the sparse user interaction area; the difference characteristics include the difference between the theoretical association strength and the actual association strength; Based on the aforementioned differences, optimization suggestions are generated.

25. A calibration device for a heat map, comprising: At least one processor; as well as, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enable the at least one processor to: Obtain the current behavior trajectory and complete behavior trajectory of the session on the same page; the current behavior trajectory is a sequence of user behavior data up to the time of the preset data entry point reporting. Based on the complete behavioral trajectory of the user terminal, determine the adjacent subsequent behaviors of the current behavioral trajectory; By combining the final behavior of the current behavior trajectory with the adjacent subsequent behavior, the state probability of each page state is determined, and the page state corresponding to the current behavior trajectory is calibrated. A page heatmap is generated based on the calibrated page status and the user behavior data.

Images