A real-time analysis system for enterprise business data based on mobile BI

Abstract

The present invention relates to the field of data analysis technology, and discloses a real-time analysis system for enterprise business data based on mobile BI, comprising: a data ingestion module for receiving enterprise business data and writing it into a cloud memory table to obtain a sequence of events to be processed; an event cluster stability window identification module for clustering the sequence of events to be processed and generating an event cluster stability window; an indicator calculation module for calculating total sales, total customer traffic, and total number of units of merchandise in stock to generate an indicator data packet; a redundant token removal module for generating tokens based on the indicator data packet, performing stack comparison to identify redundant shadows, transferring them to a black hole pool for regular cleaning, and outputting a valid indicator data packet; and a mutation color scale extraction module for associating text annotations with valid indicator data packets, extracting emotional keywords, calculating semantic situation mutation color scales, and obtaining a colored indicator data packet. The present invention achieves consistent aggregation, deduplication push, and anomaly visualization of enterprise business data in seconds.

Description

Technical Field

[0001] The present invention belongs to the field of data analysis technology, and specifically relates to a real-time analysis system for enterprise business data based on mobile BI. Background Art

[0002] With the digital transformation of industries like retail and fast-moving consumer goods, businesses are increasingly demanding real-time analysis of operational data. Mobile BI platforms, as convenient data analysis tools, are required to process high-frequency data such as POS transactions, customer flow statistics, and inventory changes in real time, providing managers with instant decision support.

[0003] However, existing technologies have three major pain points: First, data disorder and delay problems. Business operating data is generated concurrently from multiple sources, which can easily lead to timestamp errors or late data, resulting in distorted indicator calculations; second, redundant data interference. The same indicator may be pushed repeatedly due to system retries and network jitter, causing frequent refreshes or even freezes on the mobile interface; third, abnormal situation identification is inefficient. Traditional systems only display numerical values, and manual comparison of historical data is required to determine anomalies, making it difficult to capture business mutations in a timely manner. Summary of the Invention

[0004] The present invention provides a real-time analysis system for enterprise business data based on mobile BI to solve technical problems in related technologies.

[0005] The present invention provides a real-time analysis system for enterprise business data based on mobile BI, comprising:

[0006] The data ingestion module is used to receive enterprise operating data and write it into the cloud memory table to obtain the sequence of events to be processed;

[0007] An event cluster stable window identification module is used to cluster the event sequence to be processed based on the time grid, generate event cluster stable windows, and assign a unique window identifier to each event cluster stable window;

[0008] The indicator calculation module is used to calculate the total sales, total customer flow, and total number of units of merchandise inventory based on the business operating data within the event cluster stability window, and generate an indicator data packet with a timestamp;

[0009] Redundant elimination token module, used to generate tokens based on indicator data packets, perform stack comparison to identify redundant shadows, transfer redundant shadows to the black hole pool and clear them regularly, and output a unique valid indicator data packet;

[0010] The mutation color scale extraction module is used to associate text annotations with valid indicator data packets, extract emotional keywords, calculate the semantic situation mutation color scale according to preset semantic mapping rules, and color the valid indicator data packets to obtain colored indicator data packets.

[0011] Furthermore, the event sequence to be processed is clustered based on the time grid to generate an event cluster stability window, including:

[0012] S201, dynamically optimizing the time grid width based on a genetic algorithm, specifically: randomly generating N time grid width candidate values ​​as an initial population, and performing the following evaluation operations for each time grid width candidate value in the population:

[0013] Divide the timestamp of the historical event sample by the candidate value and round it down, map it to the discrete time bucket number, and obtain a set of historical event samples with the time bucket number;

[0014] Calculating the fitness of the candidate value based on the historical event sample set with the time bucket number;

[0015] Perform selection, crossover, and mutation operations on the population according to fitness; when a preset number of iterations is reached, stop optimization and output the global final time grid width as the final width; round down the event timestamps in the event sequence to be processed according to the final width and map them to the time bucket number, and output the event sequence to be processed with the time bucket number;

[0016] S202, receiving the sequence of to-be-processed events with time bucket numbers outputted in S201, grouping them by the store ID and product category combination key in the time bucket events to generate candidate event clusters;

[0017] S203, counting the number of events in each candidate event cluster generated in S202, and if the number of events in the candidate event cluster exceeds a preset stability threshold, marking the candidate event cluster as an anchor event cluster, otherwise marking it as a free event cluster;

[0018] S204: Merge the anchor event clusters with consecutive time bucket numbers marked in S203 to generate an event cluster stable window with a unique window identifier.

[0019] Furthermore, the event cluster stable window identification module also includes a free event recovery mechanism, specifically: setting a waiting coefficient. When the free event exceeds the waiting coefficient multiplied by the time grid width since the end time of the time bucket to which it belongs and is still not filled, it is judged as invalid data and discarded, wherein the value range of the waiting coefficient is 1 to 2.

[0020] Furthermore, the total sales, total customer traffic, and total number of units in stock are calculated based on the business operating data, and a timestamp-carrying indicator data packet is generated, including:

[0021] S301, extracting and normalizing the sales volume, customer flow, and number of units of merchandise inventory in the stable window of the event cluster as input for indicator calculation;

[0022] S302, traversing the stable window of the event cluster to accumulate the total sales, total customer flow, and total number of units of merchandise in stock;

[0023] S303, performing uniqueness and numerical accuracy checks on the total sales, total customer traffic, and total number of units in stock, and recording abnormal flags;

[0024] S304: Determine the window timestamp and encapsulate the unique window identifier, total sales, total customer flow, and total number of merchandise inventory units to form an indicator data packet.

[0025] Furthermore, a token is generated based on the indicator data packet, and a stack comparison is performed to identify redundant shadows. The redundant shadows are transferred to the black hole pool and cleared regularly, and a unique valid indicator data packet is output, including:

[0026] S401, generating a token based on the indicator data packet and pushing the token into a token stack, wherein the token includes: a window timestamp, a unique window identifier, and an indicator domain identifier, and the indicator domain identifier includes: total sales, total customer traffic, and total number of merchandise inventory units;

[0027] S402, determining whether the current token is a redundant shadow based on the consistency between the current token and the previous token in the token stack and whether the time difference between the two tokens meets a preset condition;

[0028] S403, writing the marked redundant shadows and corresponding indicator data packets into the black hole pool, and regularly deleting the redundant shadows according to the set clearing cycle;

[0029] S404: Output the indicator data packet corresponding to the token that is not marked as a redundant shadow as the only valid indicator data packet.

[0030] Furthermore, S402 includes: comparing the hash values ​​of the current token and the previous token in the token stack, and only when the hash values ​​of the two are the same and the corresponding indicator domain identifiers are consistent, entering the time difference determination;

[0031] Calculate the difference between the current token window timestamp and the adjacent previous token timestamp. When the difference is less than or equal to the first preset threshold, mark the current token as a redundant shadow, where the first preset threshold is used to limit the maximum repeated push time interval allowed for the same window and the same indicator domain identifier.

[0032] Furthermore, text annotations are associated with the valid indicator data packets, emotional keywords are extracted, the semantic situation mutation color level is calculated according to the preset semantic mapping rules, and the valid indicator data packets are colored to obtain the colored indicator data packets, including:

[0033] S501, obtaining log text in a preset time window, segmenting the log text and extracting emotional keywords;

[0034] S502, matching the emotional keywords with the preset semantic mapping rules, and outputting the semantic situation mutation color scale;

[0035] S503, querying a preset color mapping table according to the semantic situation mutation color level, writing a valid indicator data packet, and forming a coloring indicator data packet.

[0036] Furthermore, the preset semantic mapping rules of S502 include: when the emotional keyword matches any keyword in the preset positive keyword table, the semantic situation mutation color level is set to 2; when it matches any keyword in the preset negative keyword table, the semantic situation mutation color level is set to 1; if there is no match in the positive or negative keyword table, the semantic situation mutation color level is set to 0.

[0037] The present invention also includes: a mobile terminal rendering module, which is used to sequentially receive the coloring index data packets output by the redundant token elimination module and colored by the mutation color level extraction module, and refresh and present them on the mobile terminal interface.

[0038] The beneficial effects of the present invention are as follows: the present invention realizes one-stop processing from multi-source business data of enterprises to second-level visual indicators on mobile terminals by constructing a five-level streaming link of "data intake-event cluster stable window-indicator calculation-redundancy elimination-semantic color scale"; aligns disordered events through time grid and stable threshold to improve data consistency; removes duplicate pushes through token stack + black hole pool to reduce end-side screen flashing; instantly marks red, yellow and gray through keyword semantic mapping to enhance the discoverability of anomalies; significantly improves real-time analysis accuracy, performance and explainability without the need for complex models and high-cost hardware, and is suitable for high-concurrency scenarios such as chain retail and e-commerce promotions, achieving low-cost and rapid deployment with good scalability and compatibility. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 This is a module diagram of a real-time analysis system for enterprise business data based on mobile BI of the present invention. DETAILED DESCRIPTION

[0040] The subject matter described herein will now be discussed with reference to the embodiments of the present invention. It should be understood that the discussion of these embodiments is intended solely to enable those skilled in the art to better understand and implement the subject matter described herein, and that the functions and arrangements of the elements discussed may be varied without departing from the scope of this specification. Various examples may omit, substitute, or add various processes or components as needed. In addition, features described with respect to some examples may also be combined in other examples.

[0041] It should be noted that, unless otherwise defined, the technical or scientific terms used in one or more embodiments of the present invention should have the usual meanings understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in one or more embodiments of the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprising" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, but do not exclude other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative positional relationships. When the absolute position of the object being described changes, the relative positional relationship may also change accordingly.

[0042] like Figure 1 As shown, a real-time analysis system for enterprise business data based on mobile BI includes:

[0043] The data ingestion module 101 is used to receive enterprise operation data and write it into the cloud memory table to obtain a sequence of events to be processed;

[0044] An event cluster stable window identification module 102 is configured to cluster the event sequence to be processed based on a time grid, generate event cluster stable windows, and assign a unique window identifier to each event cluster stable window;

[0045] The indicator calculation module 103 is used to calculate the total sales, total customer flow and total number of units of merchandise in stock based on the business data within the event cluster stability window, and generate an indicator data packet with a timestamp;

[0046] Redundant token removal module 104, used to generate tokens based on indicator data packets, perform stack comparison to identify redundant shadows, transfer redundant shadows to a black hole pool and clear them regularly, and output a unique valid indicator data packet;

[0047] The mutation color scale extraction module 105 is used to associate text annotations with valid indicator data packets, extract emotional keywords, calculate semantic situation mutation color scales according to preset semantic mapping rules, and color the valid indicator data packets to obtain colored indicator data packets.

[0048] In one embodiment of the present invention, the business operation data includes but is not limited to POS transaction records, store customer flow statistics, commodity inventory change information, etc.; for the received business operation data, the data intake module first performs a data cleaning operation, fills the missing fields with reasonable data, marks the format error data with an abnormal identifier, and temporarily stores it in the abnormal data buffer, and manual verification determines whether to include it in the processing flow; duplicate data is deduplicated using the transaction serial number; and the cleaned data is written to the cloud memory table.

[0049] Once data is successfully written to the cloud memory table, the data ingestion module triggers the event generation mechanism, converting each piece of data into a pending event containing a unique event ID, timestamp, data content, and source identifier. These events are then sorted in ascending timestamp order to form a pending event sequence. This sequence is then pushed to the event cluster stable window identification module via the memory queue, completing the data flow between modules.

[0050] In one embodiment of the present invention, clustering the event sequence to be processed based on the time grid to generate an event cluster stability window includes:

[0051] S201, dynamically optimizing the time grid width based on a genetic algorithm, specifically: randomly generating N time grid width candidate values ​​as an initial population, and performing the following evaluation operations for each time grid width candidate value in the population:

[0052] Divide the timestamp of the historical event sample by the candidate value and round it down, map it to the discrete time bucket number, and obtain a set of historical event samples with the time bucket number;

[0053] Based on the set of historical event samples with time bucket numbers, the fitness of the candidate value is calculated; the fitness calculation is based on a weighted combination of the information entropy of the event distribution within the time bucket and the dispersion of the anchor candidate cluster distribution;

[0054] Perform selection, crossover, and mutation operations on the population according to fitness; when a preset number of iterations is reached, stop optimization and output the global final time grid width as the final width; round down the event timestamps in the event sequence to be processed according to the final width and map them to the time bucket number, and output the event sequence to be processed with the time bucket number;

[0055] S202, receiving the sequence of to-be-processed events with time bucket numbers outputted in S201, grouping them by the store ID and product category combination key in the time bucket events to generate candidate event clusters;

[0056] S203, counting the number of events in each candidate event cluster generated in S202, and if the number of events in the candidate event cluster exceeds a preset stability threshold, marking the candidate event cluster as an anchor event cluster, otherwise marking it as a free event cluster;

[0057] S204: Merge the anchor event clusters with consecutive time bucket numbers marked in S203 to generate an event cluster stable window with a unique window identifier.

[0058] S201: Dynamic optimization of time grid width based on genetic algorithm

[0059] Specifically, the above S201 may be implemented as follows:

[0060] Initialize the population and randomly generate N candidate values ​​for the time grid width (for example, randomly select values ​​from 1 minute to 60 minutes) to form the initial population.

[0061] Time bucket mapping, for each candidate width, the timestamp of the historical event sample is divided by the width and rounded down to obtain the corresponding time bucket number. For example, if the width is 5 minutes, the event with a timestamp of 10:03 will be mapped to time bucket number 2 (10:00-10:05).

[0062] Fitness calculation:

[0063] Information entropy calculation: statistics on the distribution of the number of events in each time bucket, and calculate the information entropy. The lower the information entropy, the more concentrated the distribution of events in the time bucket;

[0064] Specifically, the calculation formula of information entropy is: ,in, represents information entropy, T represents the total number of time buckets, t represents the time bucket index, represents the number of events in the tth time bucket, and N represents the total number of events in all time buckets.

[0065] The distribution dispersion of anchor candidate clusters identifies candidate clusters that may become anchor points (for example, time buckets where the number of events exceeds a threshold) and calculates the distribution dispersion of these candidate clusters on the time axis. Lower dispersion indicates a denser candidate cluster. Specifically, a threshold of 5 is set to filter out the time bucket sequence of anchor candidate clusters. The standard deviation of the sequence is then calculated to measure its distribution dispersion on the time axis.

[0066] Weighted combination, the information entropy and distribution dispersion are weighted summed to obtain the fitness value; specifically, the information entropy and distribution dispersion are first normalized using the maximum and minimum normalization method, and then weighted summed using preset weights to obtain the fitness value. The fitness value calculation formula is: , where F represents the fitness value, represents the information entropy weight, represents the dispersion weight, and They represent the normalized information entropy and distribution dispersion respectively.

[0067] The fitness calculation in this embodiment normalizes and weights the information entropy of time bucket events and the discreteness of the anchor point cluster distribution, which can simultaneously measure "local peak concentration" and "global anchor point compactness" to avoid single-indicator bias; the genetic algorithm thus quickly converges to the optimal grid width, making the stable window both complete and coherent and in line with the business rhythm, providing higher-quality input for subsequent indicator statistics and anomaly detection.

[0068] Genetic Operations:

[0069] Selection, select excellent individuals to enter the next generation based on fitness value.

[0070] Crossover: Randomly select two individuals for crossover to generate new candidate widths.

[0071] Mutation: mutate some individuals and introduce new candidate widths.

[0072] Iterative optimization repeats the above steps until the preset number of iterations is reached and the optimal time grid width is output.

[0073] Map pending events, using the final width to map the sequence of pending events to time buckets.

[0074] The present invention automatically searches for the optimal time grid width through a genetic algorithm, avoids the subjectivity of manual setting, and can adapt to event distribution characteristics in different business scenarios.

[0075] The specific calculation process of S202 is as follows:

[0076] Extract the composite key: extract the store ID and product category from each event to form a composite key (for example, "Store A - Clothing").

[0077] Grouping and aggregation: Events with the same composite key within the same time bucket are grouped together to form a candidate event cluster. For example, within the time bucket 10:00-10:05, all "Store A - Clothing" events are aggregated into one candidate event cluster.

[0078] The present invention groups related events by store ID and product category, laying the foundation for subsequent identification of stable event patterns and reducing interference from irrelevant events.

[0079] The specific implementation process of the above S203 is as follows:

[0080] Count the number of events. For each candidate event cluster, count the number of events contained in it.

[0081] Threshold judgment compares the number of events with a preset stability threshold. For example, if the threshold is 10, candidate event clusters with more than 10 events are marked as anchor event clusters, otherwise they are marked as free event clusters.

[0082] The present invention can identify event clusters (anchor points) with stable business patterns through threshold screening, filter out accidental events (free events), and improve the pertinence and reliability of subsequent analysis.

[0083] The specific implementation process of the above S204 is as follows:

[0084] To sort and identify consecutive events, sort the anchor event clusters by time bucket number and identify those with consecutive time bucket numbers. For example, the anchor event clusters for time buckets 10:00-10:05, 10:05-10:10, and 10:10-10:15 are considered consecutive if they belong to the same composite key.

[0085] Window merging combines consecutive anchor event clusters into a stable event cluster window and assigns a unique identifier to each window. For example, the anchor event clusters of the three consecutive time buckets above are merged into a window with a time range of 10:00-10:15.

[0086] By merging continuous anchor event clusters, the present invention can form a complete business stability period window, which is convenient for subsequent continuous analysis and monitoring of business models and captures long-term stable business trends.

[0087] In one embodiment of the present invention, the event cluster stable window identification module also includes a free event recovery mechanism. Specifically, the free event recovery mechanism uses a waiting coefficient as a core parameter. The value range of the waiting coefficient is 1 to 2, and it can be dynamically adjusted according to the real-time requirements of the business scenario. When the mechanism is started, the end time of the time bucket to which the free event belongs is first obtained. The end time is calculated by the time grid width and the start timestamp of the time bucket. For free events, the module continuously monitors its completion status. Completion refers to the candidate event cluster to which the free event belongs adding new events in the subsequent time bucket, so that the total number of events reaches a preset stability threshold, thereby becoming an anchor event. When the duration of the free event from the end time of the time bucket to which it belongs exceeds the threshold of the waiting coefficient multiplied by the time grid width, if the completion is still not completed, the free event is determined to be invalid data and discarded.

[0088] Through the above-mentioned free event recovery mechanism, the present invention can not only reserve a reasonable waiting time for potentially fillable free events, but also promptly clean up invalid free events, avoiding processing delays caused by the accumulation of redundant data, and further improving the reliability of the event cluster stable window and system operation efficiency.

[0089] In one embodiment of the present invention, total sales, total customer traffic, and total merchandise inventory units are calculated based on business operating data, and an indicator data packet with a timestamp is generated, including:

[0090] S301, extracting and normalizing the sales volume, customer flow, and number of units of merchandise inventory in the stable window of the event cluster as input for indicator calculation;

[0091] Specifically, in the current step, the indicator calculation module first extracts raw data fields related to sales, customer flow, and the number of units of merchandise in stock from the event cluster stability window and performs standardization processing. The raw data fields include, but are not limited to, "transaction amount," "transaction time," and "product code" in POS transaction records, and "number of store visitors" and "time period identifier" in store customer flow statistics. Standardization processing includes converting the monetary units of different data sources into base currency units, unifying the customer flow count format to integers, and normalizing product codes by removing special characters and completing digits to ensure the accumulativeness of indicators across data sources.

[0092] S302, traversing the stable window of the event cluster to accumulate the total sales, total customer flow, and total number of units of merchandise in stock;

[0093] S303: Perform uniqueness and numerical accuracy checks on total sales, total customer traffic, and total merchandise inventory units, and record exception flags. The uniqueness check is performed by comparing multiple cumulative results of the same indicator within the window. If the multiple calculated results are inconsistent, a "uniqueness exception" is marked. The numerical accuracy check performs format verification on total sales, total customer traffic, and total merchandise inventory units. If the format requirements are not met, a "precision exception" is marked.

[0094] S304, determining the window timestamp, and encapsulating the unique window identifier, total sales, total customer flow, and total number of merchandise inventory units to form an indicator data packet; wherein the indicator data packet adopts JSON format.

[0095] Through the above process, the indicator calculation module realizes the transformation from raw event data to standardized indicators, ensures the accuracy and consistency of the indicators through cumulative calculation and verification, and the generated indicator data packet provides structured and reliable input for the system's subsequent redundancy elimination and semantic situation analysis.

[0096] In one embodiment of the present invention, a token is generated based on an indicator data packet, a stack comparison is performed to identify redundant shadows, the redundant shadows are transferred to a black hole pool and periodically cleared, and a unique valid indicator data packet is output, including:

[0097] S401: Generate a token based on the indicator data packet and push the token into a token stack. The token includes a window timestamp, a unique window identifier, and an indicator domain identifier. The indicator domain identifier includes total sales, total customer traffic, and total number of merchandise inventory units. The token is a unique identifier carrier for the indicator data packet. The indicator domain identifier is used to distinguish different types of business indicators, including three dimensions: total sales, total customer traffic, and total number of merchandise inventory units. Each dimension corresponds to a unique domain code. After the token is generated, it is immediately pushed to the top of the token stack.

[0098] S402, determining whether the current token is a redundant shadow based on the consistency between the current token and the previous token in the token stack and whether the time difference between the two tokens meets a preset condition;

[0099] Specifically, the hash values ​​of the current token and the previous adjacent token in the token stack are compared. Only when the hash values ​​of the two are the same and the corresponding indicator domain identifiers are consistent, the time difference determination is entered; the difference between the current token window timestamp and the previous adjacent token timestamp is calculated. When the difference is less than or equal to a first preset threshold, the current token is marked as a redundant shadow, where the first preset threshold is used to limit the maximum repeated push time interval allowed for the same window and the same indicator domain identifier;

[0100] S403: Write the marked redundant shadows and corresponding indicator data packets to the black hole pool and regularly delete the redundant shadows according to the set clearing cycle. The black hole pool represents a temporary storage area for redundant data. The deletion operation completely deletes the redundant shadows and related indicator data packets in the black hole pool, freeing up system storage resources.

[0101] S404: Output the indicator data packet corresponding to the token that is not marked as a redundant shadow as the only valid indicator data packet.

[0102] Through the above process, the redundant token elimination module realizes the accurate identification and efficient filtering of duplicate indicator data packets. It ensures the accuracy of redundant judgment through the timing comparison of the token stack, and optimizes the system resource usage through the timed cleanup mechanism of the black hole pool. The final output of the valid indicator data packet provides a unique and reliable input source for subsequent modules.

[0103] In one embodiment of the present invention, text annotations are associated with valid indicator data packets, emotional keywords are extracted, semantic situation mutation color levels are calculated according to preset semantic mapping rules, and the valid indicator data packets are colored to obtain colored indicator data packets, including:

[0104] S501, obtaining log text in a preset time window, segmenting the log text and extracting emotional keywords; wherein the log text sources include but are not limited to POS system operation logs, customer service feedback records, inventory adjustment notes, etc. The word segmentation process uses a dictionary-based forward maximum matching algorithm to split the log text into independent words, such as "Today, store D's sales surged, inventory is running out" is split into "Today | Store D | Sales | Surge |, | Inventory | Emergency"; emotional keyword extraction is performed by comparing the word segmentation results with a preset emotional vocabulary to select words that meet emotional characteristics as emotional keywords, for example, extracting emotional keywords such as "surge" and "Emergency" from the log file;

[0105] S502, matching the emotional keywords with preset semantic mapping rules, and outputting a semantic situation mutation color scale; wherein the preset semantic mapping rules are set based on the semantic tendency of the emotional keywords, and include three levels of mapping relationships: positive, negative, and neutral; when the emotional keyword matches any keyword in a preset positive keyword table, the semantic situation mutation color scale is set to 2; when it matches any keyword in a preset negative keyword table, the semantic situation mutation color scale is set to 1; if no positive or negative keyword table is matched, the semantic situation mutation color scale is set to 0; wherein the positive keyword table includes at least "growth, surge, hot sales, skyrocketing", and the negative keyword table includes at least "decline, sudden drop, sluggish sales, plummeting";

[0106] S503, query the preset color mapping table according to the semantic situation mutation color level, write the valid indicator data packet, and form a coloring indicator data packet; the preset color mapping table defines the correspondence between the color level value and the visual color, for example, color level 0 corresponds to gray, indicating a steady state, color level 1 corresponds to yellow, indicating an alert state, and color level 2 corresponds to red, indicating a mutation state; the coloring indicator data packet adds a color level field to store the color code on the basis of retaining the original information of the valid indicator data packet.

[0107] Through the above process, the mutation color scale extraction module realizes the correlation analysis between text semantics and data indicators, transforming abstract business situations into intuitive color labels. The generated coloring indicator data packet provides the mobile rendering module with indicator data with color attributes that can be directly used for visual display, helping users to quickly perceive and respond to business anomalies.

[0108] In one embodiment of the present invention, it also includes: a mobile rendering module, which is used to sequentially receive the colored indicator data packets output by the redundant elimination token module and colored by the mutation color scale extraction module, and refresh and present them on the mobile terminal interface; specifically, the mobile terminal rendering module extracts the data fields in the colored indicator data packets in sequence according to the order of the colored indicator data packets in the cache queue, and maps the total sales, total customer flow and total number of merchandise inventory units to the indicator card components preset in the mobile terminal interface respectively, and directly applies the color scale color coding to the background color of the indicator card to achieve a visual display of the semantic situation.

[0109] It should be noted that the intervals and thresholds are set for ease of comparison. The threshold size depends on the amount of sample data and the cardinality set by those skilled in the art for each set of sample data, as long as it does not affect the proportional relationship between the parameter and the quantized value. Furthermore, the above formulas are all dimensionless numerical calculations. These formulas are derived from software simulations of the most recent real-world conditions using large amounts of data. The preset parameters in these formulas are set by those skilled in the art based on actual conditions.

[0110] The above describes the embodiments of the present invention, but the present invention is not limited to the above specific implementation methods. The above specific implementation methods are merely illustrative and not restrictive. Ordinary technicians in this field can also make many forms based on the inspiration of this embodiment, all of which are protected by this embodiment.

Claims

1. A real-time analysis system for enterprise business data based on mobile BI, characterized by: include: The data ingestion module is used to receive enterprise operating data and write it into the cloud memory table to obtain the sequence of events to be processed; The event cluster stable window identification module is used to cluster the event sequence to be processed based on the time grid, generate event cluster stable windows, and assign a unique window identifier to each event cluster stable window. Specifically, it includes: S201, dynamically optimizing the time grid width based on a genetic algorithm, specifically: randomly generating N time grid width candidate values ​​as an initial population, and performing the following evaluation operations for each time grid width candidate value in the population: Divide the timestamp of the historical event sample by the candidate value and round it down, map it to the discrete time bucket number, and obtain a set of historical event samples with the time bucket number; Calculating the fitness of the candidate value based on the historical event sample set with the time bucket number; Perform selection, crossover, and mutation operations on the population according to fitness; when a preset number of iterations is reached, stop optimization and output the global final time grid width as the final width; round down the event timestamps in the event sequence to be processed according to the final width and map them to the time bucket number, and output the event sequence to be processed with the time bucket number; S202, receiving the sequence of to-be-processed events with time bucket numbers outputted in S201, grouping them by the store ID and product category combination key in the time bucket events to generate candidate event clusters; S203, counting the number of events in each candidate event cluster generated in S202, and if the number of events in the candidate event cluster exceeds a preset stability threshold, marking the candidate event cluster as an anchor event cluster, otherwise marking it as a free event cluster; S204, merging the anchor event clusters with consecutive time bucket numbers marked in S203 to generate an event cluster stable window with a unique window identifier; The indicator calculation module is used to calculate the total sales, total customer traffic, and total number of units of inventory based on the company's operating data within the event cluster stability window, and generate an indicator data packet with a timestamp. Specifically, it includes: S301, extracting and normalizing the sales volume, customer flow, and number of units of merchandise inventory in the stable window of the event cluster as input for indicator calculation; S302, traversing the stable window of the event cluster to accumulate the total sales, total customer flow, and total number of units of merchandise in stock; S303, performing uniqueness and numerical accuracy checks on the total sales, total customer traffic, and total number of units in stock, and recording abnormal flags; S304, determining the window timestamp and encapsulating the unique window identifier, total sales, total customer traffic, and total number of units of merchandise in stock to form an indicator data packet; The redundant token removal module is used to generate tokens based on the indicator data packet, perform stack comparison to identify redundant shadows, transfer the redundant shadows to the black hole pool and clear them regularly, and output a unique valid indicator data packet, including: S401, generating a token based on the indicator data packet and pushing the token into a token stack, wherein the token includes: a window timestamp, a unique window identifier, and an indicator domain identifier, and the indicator domain identifier includes: total sales, total customer traffic, and total number of merchandise inventory units; S402, based on the consistency between the current token and the previous token in the token stack and whether the time difference between the two tokens meets the preset conditions, determines whether the current token is a redundant shadow, specifically including: Compare the hash values ​​of the current token and the previous token in the token stack. Only when the hash values ​​of the two are the same and the corresponding indicator domain identifiers are consistent, enter the time difference determination; Calculate the difference between the timestamp of the current token window and the timestamp of the previous token. When the difference is less than or equal to a first preset threshold, mark the current token as a redundant shadow. The first preset threshold is used to limit the maximum repeated push time interval allowed for the same indicator domain identifier in the same window. S403, writing the marked redundant shadows and corresponding indicator data packets into the black hole pool, and regularly deleting the redundant shadows according to the set clearing cycle; S404, outputting the indicator data packet corresponding to the token not marked as redundant shadow as the only valid indicator data packet; The mutation color scale extraction module is used to associate text annotations with valid indicator data packets, extract emotional keywords, calculate the semantic situation mutation color scale according to preset semantic mapping rules, and color the valid indicator data packets to obtain colored indicator data packets.

2. The real-time analysis system for enterprise business data based on mobile BI according to claim 1 is characterized in that: The event cluster stable window identification module also includes a free event recovery mechanism, specifically: setting a waiting coefficient. When a free event exceeds the waiting coefficient multiplied by the time grid width since the end time of the time bucket to which it belongs and is still not filled, it is determined to be invalid data and discarded, where the waiting coefficient has a value range of 1 to 2.

3. The real-time analysis system for enterprise business data based on mobile BI according to claim 1 is characterized in that: Associate text annotations with valid indicator data packets, extract emotional keywords, calculate the semantic situation mutation color level according to the preset semantic mapping rules, and color the valid indicator data packets to obtain colored indicator data packets, including: S501, obtaining log text in a preset time window, segmenting the log text and extracting emotional keywords; S502, matching the emotional keywords with the preset semantic mapping rules, and outputting the semantic situation mutation color scale; S503, querying a preset color mapping table according to the semantic situation mutation color level, writing a valid indicator data packet, and forming a coloring indicator data packet.

4. The real-time analysis system for enterprise business data based on mobile BI according to claim 3 is characterized in that: The preset semantic mapping rules of S502 include: when the emotional keyword matches any keyword in the preset positive keyword table, the semantic situation mutation color level is set to 2; when it matches any keyword in the preset negative keyword table, the semantic situation mutation color level is set to 1; if there is no match in the positive or negative keyword table, the semantic situation mutation color level is set to 0.

5. The real-time analysis system for enterprise business data based on mobile BI according to claim 1 is characterized in that: Also includes: The mobile rendering module is used to sequentially receive the coloring indicator data packets output by the redundant token elimination module and colored by the mutation color level extraction module, and refresh and present them on the mobile interface.

Images