Mobile application experience optimization evaluation method and system based on user behavior evaluation

Abstract

This invention discloses a mobile application experience optimization evaluation method and system based on user behavior evaluation, relating to the field of big data processing. It includes: achieving multi-dimensional analysis of application operation status by integrating a lightweight SDK and data pipeline. The system comprises five core models: a comprehensive maturity and activity evaluation model, a fine-grained user experience analysis system, a business efficiency tracking model, a stability root cause diagnosis model, and an end-to-end performance monitoring system. By collecting key indicators and user behavior data, and combining weighted statistics, pattern recognition, and baseline comparison methods, multiple scores and diagnostic reports are dynamically generated to pinpoint the specific scenarios and technical root causes of experience decline, business inefficiency, and performance degradation. The advantages of this invention are: by constructing a multi-dimensional evaluation system, combined with algorithms to accurately locate the root causes of experience problems and correlate them with business efficiency, it provides comprehensive and implementable decision support for mobile application experience optimization.

Description

Technical Field

[0001] This invention relates to the field of big data processing, and in particular to a method and system for optimizing and evaluating mobile application experience based on user behavior evaluation. Background Technology

[0002] Mobile application platforms currently face multiple challenges: on the one hand, they need to integrate the characteristics of the power industry with the standards of Internet products, build cross-domain data models and indicator systems, and develop intelligent algorithms covering anomaly detection and root cause analysis; on the other hand, they need to deeply integrate digital capabilities into existing operational processes, solve the problem of multi-source heterogeneous data governance, and establish standardized data models to improve the efficiency of problem early warning and handling.

[0003] Most current mobile application experience evaluation methods on the market are one-dimensional, focusing only on single indicators such as performance or stability, lacking comprehensive coverage of maturity, activity, user experience, and business efficiency, and failing to fully reflect the overall experience level of the application. Some methods rely on simulated scenario testing without collecting real user interaction behavior data, making it difficult to accurately capture problems such as lag and process interruption in actual use, and the root cause analysis is superficial and cannot pinpoint specific technical problems. Other methods do not link user behavior with business logs, separating experience optimization from business value, causing optimization directions to deviate from actual needs. In addition, most methods lack dynamic baselines and multi-dimensional aggregation analysis capabilities, making it impossible to adapt to the differentiated needs of different applications and organizational units, and the warnings are not timely, making it difficult to avoid the risk of experience degradation in advance. Summary of the Invention

[0004] To improve existing methods and systems, this paper provides a mobile application experience optimization evaluation method and system based on user behavior evaluation. This method constructs an evaluation system covering multiple dimensions, combines algorithms to accurately locate the root causes of experience problems and correlate them with business efficiency, and provides comprehensive and practical decision support for mobile application experience optimization.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] Mobile application experience optimization evaluation methods based on user behavior evaluation include:

[0007] Construct a multi-dimensional comprehensive evaluation model for maturity and activity, monitor and collect key operational indicator data of mobile applications within a preset period, and dynamically generate a comprehensive score of maturity and activity aggregated by different applications, organizational units and time dimensions using a preset weighted statistical algorithm.

[0008] Construct a fine-grained user experience evaluation system, collect real user interaction behavior data within the application through a lightweight SDK, perform session segmentation and event attribution analysis on the collected raw behavior data, identify the specific scenarios and technical root causes that lead to operation lag, process interruption or experience decline, and generate a user experience health report.

[0009] Build a business efficiency evaluation model, define and track key business conversion paths and efficiency indicators by linking user behavior data with backend business logs, process business event flows through real-time data pipelines, calculate the efficiency indicators of each business line, and conduct comparative analysis based on historical baselines to evaluate the operational efficiency and value realization of business functions.

[0010] A stability evaluation and root cause analysis model is constructed. Various abnormal events during application runtime are captured through a lightweight SDK, and complete error context information is recorded. Through pattern recognition and clustering algorithms, abnormal events are classified into a limited set of root causes, and the impact range and severity of each root cause on user sessions and business operations are quantified to generate stability scores and diagnostic reports.

[0011] Construct an end-to-end performance evaluation system covering the front end, network, and server end. Collect performance indicator data along the entire link from user operation to final interface rendering. Collect the above indicators through performance data collection agents and backend monitoring probes, establish a performance baseline library, calculate the deviation between the current performance indicators and the baseline in real time, identify performance degradation points, and assess their impact on user experience and business operations.

[0012] Preferably, the specific construction and execution process of the comprehensive evaluation model of maturity and activity includes:

[0013] The management backend provides a configuration interface where administrators can define specific metrics for participating in maturity and activity evaluations and assign calculation weights to each metric.

[0014] The background task periodically extracts the raw data of defined metrics from the real-time data stream and performs data cleaning, deduplication, and formatting.

[0015] Based on the preprocessed data, the original score of each evaluation object at a single time point is calculated according to the preset weighted summation scoring algorithm model. The original scores are then aggregated and calculated in a rolling manner according to the time dimension and the organizational dimension to generate the aggregated comprehensive score.

[0016] The calculated comprehensive score and key indicator trends are displayed in a graphical dashboard, and threshold alarms are set so that a notification is automatically triggered when the score falls below the preset threshold.

[0017] Preferably, the step of calculating the original score of each evaluation object at a single time point based on the preprocessed data according to a preset weighted summation scoring algorithm model, and performing rolling aggregation calculation on the original scores according to the time dimension and organizational dimension to generate the aggregated comprehensive score specifically includes:

[0018] Based on the preprocessed data, the values ​​of each indicator for each evaluation object at a specific time point are extracted and normalized. According to the preset indicator weight configuration, the scores of each normalized indicator are weighted and summed to obtain the original comprehensive score of the evaluation object at that time point.

[0019] The original scores are aggregated and calculated in rolling fashion according to daily, weekly, and monthly time windows. When calculating the daily score, the average of the original scores of all hours of that day is taken. When calculating the weekly score, the average of the daily scores of that week is taken.

[0020] After completing the time-dimensional aggregation, hierarchical aggregation is performed according to the organizational structure. When calculating the weekly score of a department, the average weekly score of all applications under that department is calculated.

[0021] The result calculated by aggregating time and organizational dimensions is defined as the aggregated comprehensive score.

[0022] Preferably, the user experience evaluation system specifically includes:

[0023] By integrating the SDK into mobile applications, user touch events, page lifecycle events, network request events, and UI rendering-related performance metrics are automatically captured.

[0024] Based on user identification and application startup and exit events, continuous user operations are divided into independent sessions. Within each session, a user behavior sequence stream is constructed according to event timestamps and logical relationships.

[0025] Key behavioral and performance data are extracted based on behavioral sequence streams, and the raw data is calculated and mapped based on preset quantification rules to obtain quantified experience indicators.

[0026] By analyzing abnormal patterns in behavioral sequences and combining them with stability data, we can pinpoint the specific pages, components, or interaction flows that cause a decline in user experience.

[0027] Preferably, the business energy efficiency evaluation model specifically includes:

[0028] Configure node events for critical business processes;

[0029] The user interface operation events collected by the client are associated with the business events generated by the backend business system through a unified session ID or business serial number.

[0030] Based on the associated event flow, the funnel conversion rate, average completion time, and failure node distribution efficiency indicators of each business path are calculated in real time.

[0031] By comparing business efficiency data across different user groups, application versions, or time periods, we can identify potential factors that affect business conversion.

[0032] Preferably, the stability evaluation and root cause analysis model specifically includes:

[0033] When the client SDK captures a serious exception, in addition to recording standard error information, it actively collects the device memory state, CPU usage, network connection information, current page stack, and recent user operation logs at the time of the exception, forming a rich snapshot of the exception scene.

[0034] After receiving an exception event, the backend service aggregates similar error stacks, merges a large number of repetitive individual events into representative problem entries, and filters out low-frequency or low-impact noise events.

[0035] Based on clustering algorithms and rule engines, the aggregated problems are analyzed to infer possible root causes, and the number of independent users, sessions, and business operations affected by the problem are counted to assess the impact.

[0036] Generate a detailed diagnostic report for each confirmed root cause and add the issue to a tracking list until the associated anomaly rate drops below baseline.

[0037] Preferably, the analysis of the aggregated problems based on clustering algorithms and rule engines to infer possible root causes specifically includes:

[0038] Receive a set of stability problem events, and extract a multi-dimensional feature vector for each event;

[0039] The feature vectors are used to calculate the feature similarity between events using the DBSCAN density clustering algorithm, and the original abnormal events with highly similar features are aggregated into several representative problem clusters.

[0040] For each problem cluster generated by clustering, the rule engine loads a preset root cause analysis rule base for in-depth analysis. After the rule engine performs matching, it infers the most likely root cause for each problem cluster.

[0041] Preferably, the end-to-end performance evaluation system specifically includes:

[0042] A unique tracking identifier is generated for each user action, and this identifier is transmitted and timestamps are recorded at each stage;

[0043] On the client side, the time spent on page loading, rendering, and script execution is collected. On the network monitoring point, the time spent on network protocols at all levels is collected. On the server side, the time spent on method execution, database query, and external call is collected internally.

[0044] Automatically learn historical performance data to establish dynamic performance baselines under different time periods and conditions. When newly collected performance data deviates significantly from the baseline, performance degradation alarms are automatically triggered.

[0045] By tracing end-to-end link data, we can show the distribution of the total time spent on a user operation at each stage and pinpoint the level at which performance bottlenecks are located.

[0046] Furthermore, a mobile application experience optimization evaluation system based on user behavior evaluation is proposed, including:

[0047] Multi-dimensional comprehensive evaluation module: Configure indicator weights, calculate and aggregate comprehensive scores of maturity and activity of different applications and organizational units over time;

[0048] User experience analysis module: Collects user interaction behavior data through a lightweight SDK, performs session segmentation and event attribution, and identifies the specific scenarios and technical root causes that lead to experience problems such as lag and interruption;

[0049] Business efficiency tracking module: It links user behavior with business logs, defines and calculates conversion rates and completion time efficiency indicators for key business paths in real time, and analyzes the value realization efficiency of business functions based on baseline comparison.

[0050] Stability monitoring and root cause analysis module: captures application runtime anomalies and their complete context, and classifies anomalies into a finite set of root causes through clustering and rule engine, quantifying their impact scope and level;

[0051] End-to-end performance monitoring module: Through end-to-end tracking and identification, it collects and integrates performance data from the client, network, and server, establishes a dynamic baseline library, and identifies performance degradation points and locates bottlenecks in real time;

[0052] Data Acquisition and Processing Engine: Collects raw data from multiple sources through SDK and probes, and performs cleaning, formatting, correlation and real-time stream processing to provide standardized data for upper-level analysis modules.

[0053] Compared with the prior art, the advantages of the present invention are:

[0054] A multi-dimensional evaluation system is constructed, covering five core dimensions: maturity and activity, user experience, business efficiency, stability, and end-to-end performance. This breaks through the limitations of single-indicator evaluation and enables comprehensive control over application experience. Leveraging a lightweight SDK and probe technology, it collects real user interactions and end-to-end data. Combined with session segmentation, event attribution, and clustering algorithms, it can accurately pinpoint the specific scenarios and technical root causes of issues such as lag and anomalies, avoiding blind optimization. Simultaneously, it correlates user behavior with business logs to quantify business conversion and efficiency, ensuring that experience optimization and business value enhancement are synchronized. Through dynamic baselines, threshold alerts, and visualization, it can monitor experience changes in real time, provide early warnings of risks, and support multi-dimensional aggregation analysis to adapt to the differentiated needs of different applications and organizations, providing a scientific and actionable basis for mobile application experience optimization. Attached Figure Description

[0055] Figure 1 This is a schematic diagram of the mobile application experience optimization evaluation method based on user behavior evaluation proposed in this invention;

[0056] Figure 2 This is a schematic diagram of the mobile application experience optimization evaluation system based on user behavior evaluation proposed in this invention. Detailed Implementation

[0057] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0058] A mobile application experience optimization evaluation system based on user behavior evaluation includes:

[0059] Multi-dimensional comprehensive evaluation module: Configure indicator weights, calculate and aggregate comprehensive scores of maturity and activity of different applications and organizational units over time;

[0060] User experience analysis module: Collects user interaction behavior data through a lightweight SDK, performs session segmentation and event attribution, and identifies the specific scenarios and technical root causes that lead to experience problems such as lag and interruption;

[0061] Business efficiency tracking module: It links user behavior with business logs, defines and calculates conversion rates and completion time efficiency indicators for key business paths in real time, and analyzes the value realization efficiency of business functions based on baseline comparison.

[0062] Stability monitoring and root cause analysis module: captures application runtime anomalies and their complete context, and classifies anomalies into a finite set of root causes through clustering and rule engine, quantifying their impact scope and level;

[0063] End-to-end performance monitoring module: Through end-to-end tracking and identification, it collects and integrates performance data from the client, network, and server, establishes a dynamic baseline library, and identifies performance degradation points and locates bottlenecks in real time;

[0064] Data Acquisition and Processing Engine: Collects raw data from multiple sources through SDK and probes, and performs cleaning, formatting, correlation and real-time stream processing to provide standardized data for upper-level analysis modules.

[0065] See Figure 1 As shown, the mobile application experience optimization evaluation method based on user behavior evaluation includes:

[0066] S1: Construct a multi-dimensional comprehensive evaluation model for maturity and activity, monitor and collect key operational indicator data of mobile applications within a preset period, and dynamically generate a comprehensive score of maturity and activity aggregated by different applications, organizational units and time dimensions using a preset weighted statistical algorithm.

[0067] S2: Construct a fine-grained user experience evaluation system. Collect real user interaction data within the application through a lightweight SDK. Perform session segmentation and event attribution analysis on the collected raw behavior data to identify the specific scenarios and technical root causes that lead to operation lag, process interruption or experience decline, and generate a user experience health report.

[0068] S3: Build a business efficiency evaluation model. By linking user behavior data with backend business logs, define and track key business conversion paths and efficiency indicators. Process business event streams through real-time data pipelines, calculate the efficiency indicators of each business line, and conduct comparative analysis based on historical baselines to evaluate the operational efficiency and value realization of business functions.

[0069] S4: Construct a stability evaluation and root cause analysis model. Capture various abnormal events during application runtime using a lightweight SDK, record complete error context information, classify abnormal events into a limited set of root causes using pattern recognition and clustering algorithms, quantify the impact range and severity of each root cause on user sessions and business operations, and generate stability scores and diagnostic reports.

[0070] S5: Construct an end-to-end performance evaluation system covering the front end, network, and server end. Collect performance indicator data along the entire link from user operation to final interface rendering. Collect the above indicators through performance data collection agent and backend monitoring probe, establish a performance baseline library, calculate the deviation between the current performance indicators and the baseline in real time, identify performance degradation points, and assess their impact on user experience and business operations.

[0071] In S1, a graphical configuration interface is provided in the management backend. Users can select or define new metrics from a pre-set metric library to serve as dimensions for the comprehensive evaluation of maturity and activity. Specific calculation logic and activation conditions are configured for each selected metric. An initial weight percentage is assigned to each metric. Multiple weighting schemes can be preset according to different stages of business development, and can be dynamically switched or automatically adjusted according to plan. Health thresholds are set for the final comprehensive score or key indicators, and notification rules are configured to be triggered when the score falls below the threshold.

[0072] By integrating multiple data sources such as mobile application SDKs, backend business system logs, and user behavior log platforms, raw indicator data streams are received in real-time or on a scheduled basis. A background scheduled task is configured to automatically extract raw data for various defined indicators from the data warehouse or data lake. A series of preprocessing operations are performed on the extracted raw data, including: removing invalid or duplicate data; handling missing values; standardizing data formats from different sources; and verifying the rationality and legality of the data according to business rules. The entire data preprocessing process is logged to monitor data quality and record any abnormal data fluctuations, ensuring the reliability and consistency of the data input to the scoring model.

[0073] Since the dimensions and numerical ranges of each indicator are different, the preprocessed values ​​of each indicator are first converted into standardized scores between 0 and 100 through min-max normalization and Z-score standardization. Based on the configured weights, the standardized scores of all indicators for a single evaluation object within a single calculation period are weighted and summed to calculate the object's original maturity and activity comprehensive score at that point in time.

[0074] Time-based aggregation is based on the original scores and automatically calculates higher-level time-based aggregate scores as needed. For example, when calculating a "daily score," the average or weighted average of all hourly scores for that day can be used; the same applies to calculating "weekly scores" and "monthly scores." Algorithms such as moving averages are supported to smooth out short-term fluctuations. Organizational-based aggregation: Based on the organizational structure, hierarchical aggregation is performed from bottom to top. For example, to calculate the score of a department, the scores of all applications under that department can be calculated as a weighted average based on application importance or user volume.

[0075] In S2, the session segmentation rules for constructing session and behavior sequences include: Main rule: A single session is considered to be from "application launch" to "application going to the background and exceeding the timeout threshold." Supplementary rule: If the application is woken up again after exceeding the threshold in the background, a new session is started. Each session has a unique ID, associated with the anonymous user ID, start and end times, and environment context. Behavior sequence flow construction includes: Event standardization: Unifying various collected raw events into standardized event objects, including: event ID, session ID, timestamp, event type, event attributes, and page context. Serialization and storage: Within a session, all events are strictly ordered by timestamp to form an ordered event flow.

[0076] Experience metrics are calculated from behavioral sequences to quantified scores. Based on the behavioral sequence stream, the following core metrics are calculated:

[0077] Average task completion step length: Define standard task steps for the core conversion path. Calculate the average number of events / pages triggered by users completing the task. Shorter step lengths result in higher efficiency. Task abandonment rate: The proportion of users who leave the page or quit midway through the task. Page interactive time achievement rate: Statistically calculate the proportion of all pages loading with a TTI less than a set threshold. Slow frame rate / stuttering rate: Calculate the proportion of frames with FPS below the threshold out of the total number of frames, or the number of consecutive frame drops exceeding a certain duration. Scrolling smoothness: The frequency and severity of frame drop events during list scrolling. Interface responsiveness: Click response latency: The time difference between the occurrence of a touch event and the application triggering the onClick callback. Calculate the 95th percentile value. Request success rate and time consumption: The success rate of API requests. Subjective satisfaction: In-app rating: The collection rate and average score of ratings from pop-up invitations after key processes are completed. Overall health score: Assign weights to the above key indicators and calculate a weighted "User Experience Health Score" from 0-100 for macro-level measurement and trend tracking.

[0078] The abnormal pattern recognition rule base includes: Frequent back / exit: Triggering "back" or "exit" operations more than N times consecutively on the same page within a short period of time is identified as suspected operational confusion or interface obstruction. Prolonged inactivity: On non-reading / video playback pages, remaining on the page for more than a threshold without any interactive events is identified as suspected process interruption or loading failure. Continuous operation failures: Multiple consecutive failures of the same network request, or repeated clicking of the same button with no response.

[0079] Upon identifying the above abnormal behavior patterns, immediately correlate stability data with current performance data. Root cause attribution includes: Technical root causes: For example, frequent return patterns + NullPointerException error logs on the current page = root cause: page code defects causing rendering failure. Continuous operation failures + network request timeouts and the network type at the time was weak = root cause: requests in a weak network environment were not optimized or downgraded. Product / design root causes: For example, path deviation + no relevant error logs = root cause: unclear interface guidance or unreasonable process design.

[0080] In S3, the core closed-loop paths that directly impact business objectives are identified. A unique business line ID is defined for each business line or process, key nodes and events are defined, and each business process is broken down into a series of key nodes. Take e-commerce ordering as an example:

[0081] Front-end events: Entering the product details page, clicking "Buy Now", submitting the order (completing the information), and clicking the "Pay" button;

[0082] Backend business events: order creation successful, payment request initiated, payment success callback, order status updated to "paid";

[0083] Define standardized event codes and event attributes for each event, for example:

[0084] Event: event_code: "order_created"

[0085] Attributes: {"order_id": "123456", "total_amount": 299.00, "item_count": 2, "payment_method": "alipay"}

[0086] Set performance indicators and goals:

[0087] Conversion rate metrics: Define the core funnel for each process, and calculate the conversion rate from each step to the next and the overall conversion rate;

[0088] Efficiency metrics: Define the average / median completion time of a process or step;

[0089] Quality and anomaly indicators: Define business anomaly events, such as payment failure, risk control interception, and insufficient inventory, and calculate their frequency of occurrence;

[0090] Set baselines: Based on historical data or business objectives, set healthy baselines or target values ​​for each indicator as evaluation benchmarks;

[0091] The integrated, seamless SDK captures basic interaction events. For critical business operations, such as "submitting an order," the SDK must automatically or through minimal code-based event tracking to report custom events with business semantics. During reporting, a globally unique tracking ID or session ID must be generated or transmitted.

[0092] At critical nodes in the business logic, business events are sent via log entries or message queues. These events must contain the same identifier associated with the front-end event, and should also include a timestamp, business result status, and key business attributes. The data pipeline consumes both front-end and back-end event streams in real time, merging events from different sources into the same business session or user journey. The output is an enhanced sequence of behaviors ordered by time, incorporating interaction and business status.

[0093] Based on the correlated event stream, real-time and batch metric calculations are performed. A stream processing framework is used to define business metric calculation tasks. The occurrence of specific event sequences is monitored using user ID and tracking ID as keys. The time difference between two correlated events is calculated, and the distribution is updated in real time. A sliding window count is performed on abnormal business events. The aggregated metrics calculated in real time are written to a time-series database. The raw, correlated event details are stored in a data lake or big data warehouse.

[0094] In S4, exception events reported by clients are received, parsed, decrypted, and formatted for standardization, and then stored in the original event details table. A stack fingerprint algorithm is used to merge similar errors. This algorithm typically includes: removing volatile information such as line numbers and memory addresses from the stack, identifying the top few frames of code belonging to the application itself, hashing the pruned and extracted stack information to generate a unique fingerprint, and aggregating original events with the same or highly similar fingerprints into a single issue signature. Each signature corresponds to a specific error pattern.

[0095] Deep aggregation extracts multi-dimensional features from all events under each aggregated issue signature, constructing feature vectors. Features include: Stack common features: similarity of key stack frames; Context features: distribution of operating system version, device type, application version, and network type at the time of occurrence; Temporal features: time-series patterns of event occurrence; Unsupervised clustering: the DBSCAN algorithm is used to cluster the feature vectors.

[0096] The algorithm calculates the density of event features in space and divides the set of densely connected problem signature events into problem clusters. A cluster may contain multiple problem signatures that differ slightly on the surface stack but share the same root cause. Several problem clusters are output, each representing a potential, more macroscopic root cause. Meanwhile, occasional, low-frequency events marked as noise are filtered out.

[0097] Maintain a pre-defined root cause analysis rule base, with rules defined by domain experts. The rule engine matches the characteristics of each "problem cluster". Typical rule example:

[0098] Rule 1 (Memory Issues): IF (Error type is OutOfMemoryError OR Memory usage > 95% when the exception occurs) AND (Intra-cluster events are highly concentrated on a particular machine type) THEN Root cause: Memory leaks or large image loading issues on a specific machine type / system version.

[0099] Rule 2 (Compatibility Issues): IF (The stack involves a specific system API) AND (The operating system versions are concentrated in a small version range) THEN Root Cause: Compatibility issues with a specific system API.

[0100] Rule 3 (Network Issues): IF (Exception accompanied by a large number of network request failures) AND (Network type is mostly mobile network) AND (Stack Trace involves request retry logic) THEN Root cause: Poor request strategy in weak network environment leads to chain failures.

[0101] Rule 4 (Data Issues): IF (Stack Trace involves JSON parsing / database operations) AND (Specific fields in the associated business data snapshot are frequently null or have abnormal formats) THEN Root Cause: The backend API returns data in a format that does not meet the client's expectations.

[0102] Rule 5 (Concurrency Issues): IF (Stack Trace Involves Multi-Threaded Access to Shared Resources) AND (Frequent Rapid Clicks in User Operation Tracking) THEN Root Cause: Concurrent Access Conflicts of UI Controls.

[0103] Quantitative analysis of the impact: Breadth: Number of affected users / devices: Count the number of unique users or devices that triggered the anomaly in this issue cluster. Percentage of affected sessions: Calculate the percentage of sessions experiencing this anomaly out of the total active sessions. Depth: Business failure rate: Calculate the percentage of critical business operations that failed due to this anomaly. User churn risk: Analyze the retention or churn rate changes of users who experienced this anomaly over a subsequent period. Trend: Statistically analyze the daily / weekly trend of the number of events in this issue cluster to determine whether it is a new issue, an ongoing issue, or a fixed issue.

[0104] In S5, static threshold alarms are prone to false alarms or missed alarms when dealing with massive and fluctuating performance data. Therefore, the system needs to introduce a dynamic baseline as an intelligent layer. The system continuously learns from historical data and aggregates and calculates statistical baselines for performance indicators by application version, machine type, network type, region, time period, and other dimensions. This baseline library is regularly updated to adapt to business growth and pattern changes. When new data is reported, the system automatically matches it with the corresponding historical baseline for comparison and uses statistical process control and other methods to identify genuine performance degradation rather than normal fluctuations. The alarms triggered by this will be more targeted and can be correlated with possible recent changes, achieving intelligent diagnosis that identifies signals from noise.

[0105] When an alarm is triggered or a specific slow request needs to be analyzed, the system enters a deep diagnostic phase. By tracking IDs, logs and metrics scattered across various locations are aggregated into a complete call chain, which includes multiple spans organized by time sequence and call relationships, clearly showing the complete path from client click to server-side internal call, and then to database return.

[0106] Flame graphs or call chain trees visually represent the time consumption distribution in terms of width and hierarchy, allowing you to easily pinpoint the widest flame—the performance bottleneck. Service dependency topology graphs provide a macro-level view of the call health between microservices, quickly identifying weak points. Performance hotspot rankings offer priority guidance for optimization efforts.

[0107] Once a specific slow method or slow SQL query is identified, the system automatically correlates it with relevant server resource metrics, error logs, and business events within that time period. For example, if a payment interface slows down, and the call chain shows that the time consumption is concentrated in a certain risk control query, further correlation with database monitoring reveals that the SQL query is performing a full table scan, with the root cause possibly being missing indexes or sudden changes in data volume. This multi-source data correlation analysis deeply integrates performance phenomena, technical stack bottlenecks, and operational events, greatly accelerating root cause identification and problem resolution.

[0108] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0109] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0110] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A mobile application experience optimization evaluation method based on user behavior evaluation, characterized in that, include: Construct a multi-dimensional comprehensive evaluation model for maturity and activity, monitor and collect key operational indicator data of mobile applications within a preset period, and dynamically generate a comprehensive score of maturity and activity aggregated by different applications, organizational units and time dimensions using a preset weighted statistical algorithm. Construct a fine-grained user experience evaluation system, collect real user interaction behavior data within the application through a lightweight SDK, perform session segmentation and event attribution analysis on the collected raw behavior data, identify the specific scenarios and technical root causes that lead to operation lag, process interruption or experience decline, and generate a user experience health report. Build a business efficiency evaluation model, define and track key business conversion paths and efficiency indicators by linking user behavior data with backend business logs, process business event flows through real-time data pipelines, calculate the efficiency indicators of each business line, and conduct comparative analysis based on historical baselines to evaluate the operational efficiency and value realization of business functions. A stability evaluation and root cause analysis model is constructed. Various abnormal events during application runtime are captured through a lightweight SDK, and complete error context information is recorded. Through pattern recognition and clustering algorithms, abnormal events are classified into a limited set of root causes, and the impact range and severity of each root cause on user sessions and business operations are quantified to generate stability scores and diagnostic reports. Construct an end-to-end performance evaluation system covering the front end, network, and server end. Collect performance indicator data along the entire link from user operation to final interface rendering. Collect the above indicators through performance data collection agents and backend monitoring probes, establish a performance baseline library, calculate the deviation between the current performance indicators and the baseline in real time, identify performance degradation points, and assess their impact on user experience and business operations.

2. The mobile application experience optimization evaluation method based on user behavior evaluation according to claim 1, characterized in that, The specific construction and execution process of the comprehensive evaluation model for maturity and activity includes: The management backend provides a configuration interface where administrators can define specific metrics for participating in maturity and activity evaluations and assign calculation weights to each metric. The background task periodically extracts the raw data of defined metrics from the real-time data stream and performs data cleaning, deduplication, and formatting. Based on the preprocessed data, the original score of each evaluation object at a single time point is calculated according to the preset weighted summation scoring algorithm model. The original scores are then aggregated and calculated in a rolling manner according to the time dimension and the organizational dimension to generate the aggregated comprehensive score. The calculated comprehensive score and key indicator trends are displayed in a graphical dashboard, and threshold alarms are set so that a notification is automatically triggered when the score falls below the preset threshold.

3. The mobile application experience optimization evaluation method based on user behavior evaluation according to claim 2, characterized in that, Based on the preprocessed data, the original score of each evaluation object at a single time point is calculated according to a preset weighted summation scoring algorithm model. The original scores are then aggregated in a rolling manner according to time and organizational dimensions to generate the aggregated comprehensive score. Specifically, this includes: Based on the preprocessed data, the values ​​of each indicator for each evaluation object at a specific time point are extracted and normalized. According to the preset indicator weight configuration, the scores of each normalized indicator are weighted and summed to obtain the original comprehensive score of the evaluation object at that time point. The original scores are aggregated and calculated in rolling fashion according to daily, weekly, and monthly time windows. When calculating the daily score, the average of the original scores of all hours of that day is taken. When calculating the weekly score, the average of the daily scores of that week is taken. After completing the time-dimensional aggregation, hierarchical aggregation is performed according to the organizational structure. When calculating the weekly score of a department, the average weekly score of all applications under that department is calculated. The result calculated by aggregating time and organizational dimensions is defined as the aggregated comprehensive score.

4. The mobile application experience optimization evaluation method based on user behavior evaluation according to claim 1, characterized in that, The user experience evaluation system specifically includes: By integrating the SDK into mobile applications, user touch events, page lifecycle events, network request events, and UI rendering-related performance metrics are automatically captured. Based on user identification and application startup and exit events, continuous user operations are divided into independent sessions. Within each session, a user behavior sequence stream is constructed according to event timestamps and logical relationships. Key behavioral and performance data are extracted based on behavioral sequence streams, and the raw data is calculated and mapped based on preset quantification rules to obtain quantified experience indicators. By analyzing abnormal patterns in behavioral sequences and combining them with stability data, we can pinpoint the specific pages, components, or interaction flows that cause a decline in user experience.

5. The mobile application experience optimization evaluation method based on user behavior evaluation according to claim 1, characterized in that, The business energy efficiency evaluation model specifically includes: Configure node events for critical business processes; The user interface operation events collected by the client are associated with the business events generated by the backend business system through a unified session ID or business serial number. Based on the associated event flow, the funnel conversion rate, average completion time, and failure node distribution efficiency indicators of each business path are calculated in real time. By comparing business efficiency data across different user groups, application versions, or time periods, we can identify potential factors that affect business conversion.

6. The mobile application experience optimization evaluation method based on user behavior evaluation according to claim 1, characterized in that, The stability evaluation and root cause analysis model specifically includes: When the client SDK captures a serious exception, in addition to recording standard error information, it actively collects the device memory state, CPU usage, network connection information, current page stack, and recent user operation logs at the time of the exception, forming a rich snapshot of the exception scene. After receiving an exception event, the backend service aggregates similar error stacks, merges a large number of repetitive individual events into representative problem entries, and filters out low-frequency or low-impact noise events. Based on clustering algorithms and rule engines, the aggregated problems are analyzed to infer possible root causes, and the number of independent users, sessions, and business operations affected by the problem are counted to assess the impact. Generate a detailed diagnostic report for each identified root cause and add the issue to a tracking list until the associated anomaly rate drops below baseline.

7. The mobile application experience optimization evaluation method based on user behavior evaluation according to claim 6, characterized in that, The analysis of the aggregated problems based on clustering algorithms and rule engines, and the inference of possible root causes, specifically include: Receive a set of stability problem events, and extract a multi-dimensional feature vector for each event; The feature vectors are used to calculate the feature similarity between events using the DBSCAN density clustering algorithm, and the original abnormal events with highly similar features are aggregated into several representative problem clusters. For each problem cluster generated by clustering, the rule engine loads a preset root cause analysis rule base for in-depth analysis. After the rule engine performs matching, it infers the most likely root cause for each problem cluster.

8. The mobile application experience optimization evaluation method based on user behavior evaluation according to claim 1, characterized in that, The end-to-end performance evaluation system specifically includes: A unique tracking identifier is generated for each user action, and this identifier is transmitted and timestamps are recorded at each stage; On the client side, the time spent on page loading, rendering, and script execution is collected. On the network monitoring point, the time spent on network protocols at all levels is collected. On the server side, the time spent on method execution, database query, and external call is collected internally. Automatically learn historical performance data to establish dynamic performance baselines under different time periods and conditions. When newly collected performance data deviates significantly from the baseline, performance degradation alarms are automatically triggered. By tracing end-to-end link data, we can show the distribution of the total time spent on a user operation at each stage and pinpoint the level at which performance bottlenecks are located.

9. A mobile application experience optimization evaluation system based on user behavior evaluation, used to implement the mobile application experience optimization evaluation method based on user behavior evaluation as described in any one of claims 1-8, characterized in that, include: Multi-dimensional comprehensive evaluation module: Configure indicator weights, calculate and aggregate comprehensive scores of maturity and activity of different applications and organizational units over time; User experience analysis module: Collects user interaction behavior data through a lightweight SDK, performs session segmentation and event attribution, and identifies the specific scenarios and technical root causes that lead to experience problems such as lag and interruption; Business efficiency tracking module: It links user behavior with business logs, defines and calculates conversion rates and completion time efficiency indicators for key business paths in real time, and analyzes the value realization efficiency of business functions based on baseline comparison. Stability monitoring and root cause analysis module: captures application runtime anomalies and their complete context, and classifies anomalies into a finite set of root causes through clustering and rule engine, quantifying their impact scope and level; End-to-end performance monitoring module: Through end-to-end tracking and identification, it collects and integrates performance data from the client, network, and server, establishes a dynamic baseline library, and identifies performance degradation points and locates bottlenecks in real time; Data Acquisition and Processing Engine: Collects raw data from multiple sources through SDK and probes, and performs cleaning, formatting, correlation and real-time stream processing to provide standardized data for upper-level analysis modules.

Images