Opportunity discovery processing system fusing multi-dimensional data streams

Abstract

The embodiment of the invention provides an opportunity discovery processing system fusing multi-dimensional data streams, and relates to the technical field of marketing data processing technologies. The system comprises a multi-dimensional data stream access module used for acquiring a multi-dimensional data stream in real time; the data discretization coding module is used for generating a binary coding string for the data of each dimension; the unified context hash generation module is used for generating a unified context hash character string based on the plurality of binary coding strings corresponding to the plurality of data dimensions; the opportunity discovery engine is used for performing matching query in a historical context database based on the unified context hash character string so as to identify a marketing opportunity; and the marketing scheme generation module is used for generating a marketing solution scheme in response to the marketing opportunity. According to the invention, the problem of low opportunity decision-making efficiency is solved, and the effect of improving the opportunity decision-making efficiency is further achieved.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the field of marketing data processing, and in particular, to an opportunity discovery processing system fusing multi-dimensional data streams. BACKGROUND

[0002] In modern business decision-making, enterprises need to analyze massive, multi-dimensional data streams in real time to capture fleeting marketing opportunities. These data streams are often heterogeneous, for example, they may include numerical local traffic, weather temperature, and categorical social media sentiment, and competitor promotional activities at the same time.

[0003] The prior art often relies on complex machine learning models to make opportunity predictions when dealing with such problems. However, these models have the problems of high training cost, poor interpretability, and slow response to emerging patterns. In addition, without using pre-trained models, how to efficiently define and identify "high-value opportunities" that meet multiple complex conditions from continuous data streams in a low computational overhead and logically robust manner is still a significant technical challenge, especially query latency and computational resource consumption constitute the main performance bottleneck. SUMMARY

[0004] Embodiments of the present application provide an opportunity discovery processing system fusing multi-dimensional data streams to at least solve the problem of low efficiency of opportunity decision-making in the related art.

[0005] According to an embodiment of the present application, an opportunity discovery processing system fusing multi-dimensional data streams is provided, comprising:

[0006] A multi-dimensional data stream access module is configured to acquire multi-dimensional data streams in real time, wherein the multi-dimensional data streams comprise a plurality of data dimensions representing the state of a business environment;

[0007] A data discretization and encoding module is configured to perform a discretization and encoding operation on each data dimension in the multi-dimensional data streams to generate a corresponding binary encoding string;

[0008] A unified context hash generation module is configured to generate a unified context hash string based on the plurality of binary encoding strings corresponding to the plurality of data dimensions;

[0009] An opportunity discovery engine is configured to perform a matching query in a historical context database based on the unified context hash string to identify a marketing opportunity;

[0010] A marketing solution generation module is configured to generate a marketing solution in response to the marketing opportunity.

[0011] In an exemplary embodiment, performing a discretization encoding operation for each data dimension in the multidimensional data stream to generate a corresponding binary encoded string includes:

[0012] Determine the type of the data dimension;

[0013] When the data dimension is of the first type, the numerical data is mapped to a discrete interval according to a preset quantization boundary set, and the first binary code string corresponding to the discrete interval is obtained.

[0014] When the data dimension is of type two, the categorized data is mapped to a preset enumeration value according to a preset enumeration mapping relationship, and the second binary encoding string corresponding to the preset enumeration value is obtained.

[0015] In one exemplary embodiment, generating a unified context hash string based on the multiple binary encoded strings corresponding to the multiple data dimensions includes:

[0016] Determine the maximum encoding length among the plurality of binary encoded strings;

[0017] Starting from the first and second positions of the plurality of binary encoded strings in a preset dimensional order, the bits of each binary encoded string at the current position are extracted iteratively in positional order.

[0018] The multiple bits extracted in each iteration are concatenated according to the preset dimensional order to form an intermediate string fragment;

[0019] The intermediate string fragments generated in all iterations corresponding to the maximum encoding length are combined in sequence to form the unified context hash string.

[0020] In one exemplary embodiment, the step of performing a matching query in the historical context database based on the unified context hash string to identify marketing opportunities includes:

[0021] Use the unified context hash string as the query key to perform an exact match query in the historical context database;

[0022] If the query finds a historical context record and the historical business performance indicator associated with the historical context record is higher than a first preset threshold, then the marketing opportunity is identified as a historical recurrence opportunity.

[0023] In one exemplary embodiment, after performing an exact match query in the historical context database using the unified context hash string as the query key, the system further includes:

[0024] Extract a prefix of a preset length from the unified context hash string to obtain the context prefix string;

[0025] Using the context prefix string as a query condition, a prefix matching query is performed in the historical context database to retrieve historical context records with the same prefix;

[0026] Calculate the statistical values ​​of the historical business performance indicators associated with the historical context records;

[0027] If the statistical value is higher than the second preset threshold, the marketing opportunity will be classified as a pattern-similar opportunity.

[0028] In one exemplary embodiment, after performing an exact match query in the historical context database using the unified context hash string as the query key, the system further includes:

[0029] Obtain the historical unified context hash string within a preset time window prior to the current time point to form the current context sequence;

[0030] Determine whether the current context sequence matches a predefined value transition pattern;

[0031] If a match is found, the marketing opportunity will be classified as a state transition opportunity.

[0032] In one exemplary embodiment, generating a marketing solution in response to the marketing opportunity includes:

[0033] Select the corresponding plan template from the plan template library based on the type of marketing opportunity;

[0034] Parse the parameter placeholders contained in the scheme template;

[0035] The marketing solution is generated by filling the parameter placeholders with real-time data from the multidimensional data stream.

[0036] According to another embodiment of the present invention, an opportunity discovery processing method for fusing multidimensional data streams is provided, comprising:

[0037] Real-time acquisition of multidimensional data streams, which include multiple data dimensions characterizing the state of the business environment;

[0038] For each data dimension in the multidimensional data stream, a discretization encoding operation is performed to generate a corresponding binary encoded string;

[0039] A unified context hash string is generated based on the multiple binary encoded strings corresponding to the multiple data dimensions;

[0040] Based on the unified context hash string, a matching query is performed in the historical context database to identify marketing opportunities;

[0041] In response to the aforementioned marketing opportunities, generate marketing solutions.

[0042] According to yet another embodiment of the present invention, a computer-readable storage medium is also provided, wherein a computer program is stored therein, wherein the computer program is configured to perform the steps in the above method embodiments when executed.

[0043] According to yet another embodiment of the present invention, an electronic device is also provided, including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the steps in the above method embodiments.

[0044] This invention integrates and encodes heterogeneous multidimensional data streams into single state data, thereby transforming complex multidimensional state space queries into efficient string matching operations. This solves the technical problems of low efficiency and resource waste in opportunity decision-making in the prior art, and achieves the effect of improving the efficiency of decision-making opportunities. Attached Figure Description

[0045] Figure 1 This is a flowchart of an opportunity discovery processing method that integrates multidimensional data streams according to an embodiment of the present invention;

[0046] Figure 2 This is a structural block diagram of an opportunity discovery processing system that integrates multidimensional data streams according to an embodiment of the present invention;

[0047] Figure 3 This is a flowchart of the multidimensional data encoding and fusion process according to an embodiment of the present invention;

[0048] Figure 4 This is a schematic diagram of bit extraction and splicing generated by the unified context hash according to a specific embodiment of the present invention;

[0049] Figure 5 This is a multi-strategy query link graph for opportunity identification according to embodiments of the present invention;

[0050] Figure 6 This is a schematic diagram of market potential changes according to Embodiment 2 of the present invention. Detailed Implementation

[0051] The technical solutions of the embodiments of this application will be described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0052] In the following description, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0053] Furthermore, in this application, directional terms such as "upper," "lower," "left," and "right" may be defined relative to the orientation of the components shown in the accompanying drawings. It should be understood that these directional terms can be relative concepts, used for relative description and clarification, and may change accordingly depending on the orientation of the components in the accompanying drawings.

[0054] In this application, unless otherwise expressly specified and limited, the term "connection" should be interpreted broadly. For example, "connection" can be a fixed connection, a detachable connection, or an integral part; it can be a direct connection or an indirect connection through an intermediate medium. Furthermore, the term "coupled" can refer to an electrical connection that enables signal transmission.

[0055] As used herein, “about,” “approximately,” or “approximately” includes the stated value and the average value within an acceptable range of deviation from the given value, wherein the acceptable range of deviation is determined by a person skilled in the art taking into account the measurement under discussion and the error associated with the measurement of the given quantity (i.e., the limitations of the measurement system).

[0056] This application provides an opportunity discovery processing system that integrates multidimensional data streams. This system utilizes an innovative multidimensional context hashing mechanism to integrate and encode instantaneous, heterogeneous multidimensional data streams into single state data, thereby transforming complex multidimensional state space queries into efficient string matching operations. This solves the technical problems of decision delay and resource waste in existing technologies. Figure 2 As shown, the system includes a multi-dimensional data stream access module, a data discretization encoding module, a unified context hash generation module, an opportunity discovery engine, and a marketing plan generation module. The following will refer to... Figure 1 The present invention illustrates an opportunity discovery processing method that integrates multidimensional data streams, and elaborates in detail on the functions and workflows of each module of the system described in this application.

[0057] Example 1

[0058] S100: Real-time acquisition of multidimensional data streams, which include multiple data dimensions representing the state of the business environment.

[0059] This step is executed by the multidimensional data stream access module, which collects data in real time and in parallel from multiple heterogeneous data sources. These data dimensions together constitute a comprehensive snapshot of the current business environment. The multidimensional data stream is not a static dataset, but rather time-series data that is continuously updated at a high frequency. The multiple data dimensions may include (but are not limited to): local pedestrian traffic data, weather data, social media sentiment data, competitor dynamic data, and internal inventory data. Local pedestrian traffic data can originate from infrared sensors deployed in specific business areas, video surveillance analysis systems, or regional heat map data provided by mobile operators, and its unit is typically "persons / unit time". Weather data can be obtained by calling public meteorological service APIs, including temperature (unit: degrees Celsius), humidity (unit: %), precipitation (unit: millimeters), and weather conditions (such as sunny, cloudy, rainy, etc.). Social media sentiment data can be obtained through... Data is obtained through natural language processing analysis of social media text related to specific geographic locations or keywords, and is usually quantified as a sentiment index (e.g., a floating-point number between -1 and 1). Competitor dynamic data can be obtained from web crawlers monitoring competitors' official websites and social media accounts, recording whether they are conducting promotional activities (Boolean data) or price adjustments. Internal inventory data is obtained directly from the company's own enterprise resource planning (ERP) or inventory management system, reflecting the real-time inventory level of key products (unit: pieces). To ensure data consistency, the multi-dimensional data stream access module aligns all data streams with a unified timestamp as a benchmark, for example, in seconds or minutes, aggregating all dimensions of data at that time point into an instantaneous state vector as input for subsequent processing. In addition, this acquisition process is continuous, ensuring that the system can respond to dynamic changes in the environment in near real-time.

[0060] For example, suppose at point in time The instantaneous state vector obtained by the system For: [Local foot traffic: 1250 people / hour, Temperature: 28.5 °C, Weather: "Sunny", Social media sentiment index: 0.75, Competitor A promotion status: "No", Product SKU-XYZ inventory: 350 units], this vector A complete description of in This system continuously acquires key aspects of the business environment it focuses on; for example, the system can be configured to generate a new state vector every minute, forming a high-density data stream sequence. And so on; it should be noted that the data acquisition latency should be controlled within seconds to ensure the timeliness of decision-making. For this purpose, a buffer queue can be set up inside the data access module to smooth out any instantaneous delays or jitters that may exist from different data sources and ensure a stable supply of data streams; for example, if the weather API updates data every 5 minutes, while the pedestrian flow sensor updates every 1 minute, the system will combine the most recent valid weather data with the current pedestrian flow data to form a state vector, ensuring the logical validity of the data in terms of time.

[0061] S200: For each data dimension in the multidimensional data stream, perform a discretization encoding operation to generate the corresponding binary encoded string.

[0062] This step is performed by the data discretization and encoding module. After obtaining the original multidimensional state vector, this module performs normalization processing on each component of the vector to uniformly convert heterogeneous data of different types, dimensions, and value ranges into a standardized, fixed-length binary format. Specifically, this module first determines the type of each data dimension, which is usually divided into two main categories: numerical and categorical, and then uses different strategies to process them respectively.

[0063] For numerical data, such as temperature or pedestrian traffic, the system employs a mapping method based on preset quantization boundaries. These boundaries divide the continuous numerical space of this dimension into several non-overlapping discrete intervals. Each discrete interval is assigned a unique integer index, which is then converted into a fixed-length binary code string. The quantization boundaries have a certain impact on the encoding accuracy and sensitivity to data changes. They can be set based on the statistical distribution of historical data (e.g., using percentiles to ensure that the number of data points in each interval is approximately equal) or based on domain knowledge (e.g., for temperature, intervals can be divided based on perceived comfort or seasonal characteristics). The bit length of the binary code determines the resolution of this dimension; the longer the bit length, the more intervals can be divided, and the higher the accuracy.

[0064] For categorized data, such as weather conditions ("Sunny," "Cloudy," "Rainy") or competitor promotion status ("Yes," "No"), the system employs a method based on a pre-defined enumeration mapping relationship. The system creates a lookup table for all possible values ​​of that dimension. Each enumerated value in the table (e.g., "Sunny") is pre-assigned a unique integer index, which is then converted into a fixed-length binary encoded string. The bit length of the encoding depends on the number of possible values ​​for that dimension. , usually set to The number of bits is used to ensure that all possible categories can be uniquely represented.

[0065] Optionally, the method for determining the quantization boundary set can be based on the statistical distribution of historical data. The specific steps are as follows: First, collect all historical data points for a specific numerical dimension (e.g., "local pedestrian traffic") over a relatively long period of time (e.g., three months) to form a sample set; second, sort the sample set; then, calculate specific percentiles of the sorted sample set, such as quartiles (25%, 50%, 75%), or more refined decimals. These percentile values ​​are selected as the quantization boundary. For example, if a 4-bit encoding (i.e., 16 intervals) is to be used, the 6.25, 12.5, ..., 93.75 percentiles can be calculated as 15 cut-off points.

[0066] For example, suppose we analyze 10,000 past "local pedestrian traffic" data points and calculate that the 25th percentile is 450 people / hour, the 50th percentile is 800 people / hour, and the 75th percentile is 1300 people / hour. If we use a 2-bit code (4 intervals), the quantization boundary can be set as [450, 800, 1300], thus dividing the pedestrian traffic into four intervals: (<450), [450, 800), [800, 1300), and (>=1300). This ensures that the historical data sample size contained in each discrete interval is relatively balanced, avoiding the problem that some codes will not appear for a long time due to uneven data distribution.

[0067] Another approach is to determine the boundary set based on domain knowledge, which is suitable for numerical dimensions with clear business implications. Specifically, the system also allows business experts or operations personnel to manually set quantified boundaries based on their expertise and business logic. For example, for a restaurant, the quantified boundary for the "customer traffic" dimension might be set at key time points during lunch peaks (e.g., 11:30 AM) and dinner peaks (e.g., 6:00 PM), or based on whether customer traffic reaches the threshold for activating "fast service mode"; for the "temperature" dimension, the boundary might be set at key temperature points affecting customers' choice of cold or hot drinks, such as 25°C and 15°C, and so on.

[0068] S300: Generates a unified context hash string based on multiple binary encoded strings corresponding to multiple data dimensions.

[0069] This step is executed by the unified context hash generation module to merge multiple independent binary encoded strings of different dimensions into a single string, namely the unified context hash string. In this embodiment, a bit interleaving algorithm can be used. The principle of this algorithm is to interleave information from different dimensions at the bit level, so that adjacent bits in the final hash string are likely to come from different original dimensions. This structure has an important characteristic: if the discrete codes of two multidimensional data points are the same or similar in all dimensions, then the unified context hash strings they generate will have a long common prefix. This characteristic makes one-dimensional string prefix matching efficient in simulating proximity queries in high-dimensional space, thereby greatly improving query efficiency.

[0070] like Figure 3 As shown, the specific generation process is as follows:

[0071] First, this module determines the maximum length among all input binary encoded strings, denoted as . This maximum length defines the number of iterations required for the bit interleaving operation.

[0072] Secondly, this module follows a preset, fixed dimensional order, starting from the first iteration (position order). It begins by extracting the current bit position from each binary encoded string in each iteration. If the length of an encoded string is less than the current bit position, that dimension does not contribute bits in that iteration.

[0073] Then, all the bits extracted in this iteration are concatenated strictly according to the preset dimensional order to form an intermediate string fragment.

[0074] Finally, all The intermediate string fragments generated in each iteration are combined in sequence to form the final unified context hash string.

[0075] For example, such as Figure 4 As shown, for the aforementioned output sequence ["0110", "1011", "00", "110", "0", "100"], the dimension order is: pedestrian flow (F), temperature (T), weather (W), mood (E), promotion (P), and inventory (S).

[0076] Determine the maximum length: The lengths of the input encoded strings are 4, 4, 2, 3, 1, and 3 respectively; the maximum length is... The value is 4, therefore, 4 rounds of iteration are required.

[0077] First iteration (i=0): Concatenate to obtain the intermediate segment: "010101";

[0078] Second iteration (i=1): Concatenate to obtain the intermediate segment: "10010";

[0079] Third iteration (i=2): Concatenate to obtain the intermediate segment: "1100";

[0080] Fourth iteration (i=3): Concatenate to obtain the intermediate segment: "01";

[0081] Final combination: Combine all intermediate segments in sequence: "010101" + "10010" + "1100" + "01".

[0082] The final generated unified context hash string is "01010110010110001", and so on.

[0083] S400: Based on a unified context hash string, it performs matching queries in a historical context database to identify marketing opportunities.

[0084] This step is performed by the opportunity discovery engine. After generating a unified context hash string representing the current state, the engine uses this string as the basis for a query in a pre-built historical context database. This historical database stores a large number of unified context hash strings from the past and their corresponding business performance metrics (e.g., sales, conversion rates, etc.). The opportunity discovery engine can execute various matching query strategies at different granularities to identify different types of opportunities.

[0085] Strategy 1: Use precise matching queries to identify "historical recurrence" opportunities.

[0086] The engine uses the currently generated complete unified context hash string as the query key to search the historical database for the existence of the exact same historical record. If the query is successful and the business performance indicator associated with the historical record is higher than the preset first threshold, the system determines that the current state is a high-value "historical recurrence" opportunity.

[0087] Strategy 2: Prefix matching query to identify "pattern similarity" opportunities.

[0088] The engine extracts a prefix of a preset length from the current unified context hash string and uses it as a query condition. It then retrieves all historical records with this prefix from the database to form a set of similar states. The engine then calculates the statistical value (such as the average value) of the business performance indicators of this set. If this statistical value is higher than a preset second threshold, the system determines that the current state is a "pattern similarity" opportunity.

[0089] Strategy 3: Time series pattern analysis to identify "state transition" opportunities.

[0090] The engine caches a series of unified context hash strings generated recently to form the current context sequence. At this point, the engine determines whether this sequence matches a pattern in the predefined "value transition pattern library".

[0091] For example, a value transition pattern P1 can be defined as an ordered pair of hash prefixes ["10...", "00..."], designed to capture the critical moment of "from rain to sunshine"; suppose the opportunity discovery engine obtains the current context sequence as [..., MDCH(t-2)="1011...", MDCH(t-1)="1001...", MDCH(t)="0010..."], the engine detects that the latest hash MDCH(t) in the sequence has a prefix of "00", while its immediate preceding hash MDCH(t-1) has a prefix of "10". This transition from "10..." to "00..." matches pattern P1, therefore, the system determines that a "state transition" opportunity has been discovered, and so on.

[0092] Optionally, an optimal threshold can be found using optimization methods based on backtesting of historical data to balance the precision and recall of opportunity discovery. The specific steps are as follows: First, divide the historical dataset into a training set and a test set; second, on the training set, iterate through a range of candidate thresholds, and for each candidate threshold, simulate and run the opportunity discovery algorithm, calculating its precision and recall, and then calculate a comprehensive evaluation metric, such as the F1 score; finally, select the candidate threshold that maximizes the F1 score as the final "preset threshold," and verify its performance using the test set.

[0093] For example, if the backtest results show that the F1 score is the highest (0.86) when the "second preset threshold" is 30%, the system will set the threshold to 30%.

[0094] S500: Respond to marketing opportunities and generate marketing solutions.

[0095] This step is performed by the marketing solution generation module. After the opportunity discovery engine successfully identifies a marketing opportunity, this module is responsible for building and outputting specific, actionable response strategies. Internally, this module maintains a solution template library associated with different opportunity types. Each template is an instruction framework containing parameter placeholders (such as [discount level], [main product SKU]). Upon receiving an opportunity signal, the module first selects the matching template according to the opportunity type, then parses out the placeholders, and finally uses information from the currently acquired multidimensional data stream (such as inventory data, competitor dynamics) to populate the placeholders, generating a complete marketing solution.

[0096] Example 2

[0097] Unlike Example 1, in order to address the problem that the method in Example 1 is heavily reliant on historical data and makes it difficult to discover entirely new opportunity patterns, this example calculates the intervention point with the maximum positive market response (return) through minimal marketing intervention (cost).

[0098] In this embodiment, any instantaneous market state (defined by the multidimensional data stream) corresponds to a point in the potential energy field and has a specific market potential energy value. This potential energy value is designed as a measure of marketing difficulty or cost: the lower the potential energy, the more favorable the current environment is for marketing, or the lower the cost required to achieve the same effect. Therefore, the task of opportunity discovery shifts from "matching historical high-value patterns" to "finding a path that causes the potential energy value of the current market state point to decrease the fastest and most."

[0099] Specifically, the system and method flow described in this embodiment is as follows:

[0100] S210: Real-time acquisition of multidimensional data streams and dimensional partitioning.

[0101] This step is executed by the multi-dimensional data stream access module, which acquires data in real time from multiple dimensions, including pedestrian flow, weather, and competitor dynamics. A key difference is that this embodiment logically divides these data dimensions into a controllable dimension set and an uncontrollable dimension set.

[0102] Uncontrollable dimensions: These include external environmental factors that companies cannot directly change through their own marketing activities, such as weather conditions, temperature, general social sentiment, and promotional activities launched by competitors.

[0103] Controllable dimensions: These include variables that businesses can directly control through marketing decisions, such as the discount level of their own products, advertising channels, selection of key products, and style of promotional copy. This classification forms the basis for subsequent "gradient probing."

[0104] S220: Construct the instantaneous market potential energy field and calculate the current market potential energy value.

[0105] This step is performed by the market potential field construction module. This module receives the incoming instantaneous state vector and constructs it according to a predefined potential function. Calculate the current market status The corresponding current market potential value Preferably, the total potential energy consists of two parts: the sum of the dimensional potential energy components of each dimension, and the sum of the interactive potential energy components between each dimension.

[0106]

[0107] in:

[0108] It is a state The Middle The values ​​of each dimension.

[0109] It is the first A dimensional potential energy function is used to map the original value of a dimension to a potential energy component. The design principle of this function is: values ​​that are beneficial to marketing correspond to low potential energy, and values ​​that are detrimental correspond to high potential energy.

[0110] For example, regarding the "people flow" dimension Its potential energy function can be an inverse proportional function: ,in This is the scaling factor. To prevent a small constant with a denominator of zero, it's easy to understand that the greater the flow of people, the lower the potential energy.

[0111] Regarding the "competitor promotion status" dimension (0 represents no, 1 represents yes), its potential energy function can be a linear function: This means that when competitors promote their products, the potential energy increases.

[0112] These are elements in the synergy matrix, representing the dimension. and dimensions The impact of the interaction between them on the total potential energy.

[0113] For example, This element can be a lookup table or a submatrix; if the weather is "sunny" and the featured product is "ice cream", then the corresponding... The value might be a large negative number (e.g., -50), indicating a strong positive synergistic effect between the two, significantly lowering the overall potential energy. Conversely, if the weather is "rainy" and the main product is "ice cream," then... It could be a positive number, indicating a negative synergistic effect that would increase the total potential energy, and so on.

[0114] It is a function used to combine values ​​from two dimensions; it can be a simple product. Or other forms.

[0115] It is the weight of the potential energy components in each dimension.

[0116] Thus, the system can quantify a heterogeneous, complex, multidimensional market state into a single potential value.

[0117] S230: Perform opportunistic gradient detection.

[0118] This step is performed by the Opportunity Gradient Detection module, which iterates through one or more dimensions in the controllable set of dimensions and generates a series of possible values.

[0119] For example, for the "discount level" dimension, candidate intervention actions can be generated: [no discount, 10% off, 20% off, 30% off].

[0120] For the "Featured Products" dimension, it can generate candidate intervention actions: [SKU-A, SKU-B, SKU-C], and so on.

[0121] For each candidate intervention (or combination thereof), the chance gradient detection module performs the following operations:

[0122] Create a simulated future market state The uncontrollable dimension of this state is related to the current state. The controllable dimension is kept consistent, while the value of the current candidate intervention action is replaced.

[0123] Using the aforementioned potential energy function Calculate this simulated state The corresponding simulated market potential value .

[0124] S240: Identify and quantify the optimal intervention opportunity.

[0125] This step is performed by the opportunity quantification engine. For example... Figure 6 As shown, after generating all candidate intervention actions and their corresponding simulated market potential values, this module evaluates and sorts them.

[0126] For each candidate intervention action The engine will calculate a potential energy decrease:

[0127]

[0128] This potential energy decrease The intervention actions were directly quantified. The resulting "benefits".

[0129] The engine then compares the potential energy reduction of all candidate intervention actions and selects the one that makes... The action that achieves the maximum value is identified as the optimal intervention action. This optimal intervention action is the "marketing opportunity" discovered in this embodiment.

[0130] For example:

[0131] Assuming the current .

[0132] Candidate intervention action 1: {Discount: 10% off}, calculated as follows ,but .

[0133] Candidate intervention action 2: {Discount: 20% off}, calculated as follows ,but .

[0134] Candidate intervention action 3: {Main product: SKU-B}, calculated as follows ,but .

[0135] After comparing the opportunity quantification engines, it was found that... The maximum. Therefore, the optimal intervention action is determined to be {discount: 80%}, and so on.

[0136] S250: A marketing solution driven by generating potential energy.

[0137] This step is performed by the marketing plan generation module, whose input is the previously determined optimal intervention action. Then, this abstract action is transformed into an executable marketing instruction.

[0138] For example, if the optimal intervention action is {discount: 20% off, featured product: SKU-B}, the marketing plan generation module will generate an instruction, such as: {"command": "update_price", "target": "SKU-B", "discount": "0.8"}, and at the same time generate another instruction {"command": "update_ad_content", "target": "main_screen", "content_template": "hot_item_promo", "sku": "SKU-B"}, and so on.

[0139] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0140] It should be noted that the above modules can be implemented by software or hardware. For the latter, they can be implemented in the following ways, but are not limited to: all the above modules are located in the same processor; or, the above modules are located in different processors in any combination.

[0141] Embodiments of the present invention also provide a computer-readable storage medium storing a computer program, wherein the computer program is configured to perform the steps in any of the above method embodiments when executed.

[0142] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard disk, magnetic disk, or optical disk.

[0143] Embodiments of the present invention also provide an electronic device including a memory and a processor, the memory storing a computer program and the processor being configured to run the computer program to perform the steps in any of the above method embodiments.

[0144] In one exemplary embodiment, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor and the input / output device is connected to the processor.

[0145] Through the above description of the embodiments, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0146] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another apparatus, or some features may be ignored or not executed. Furthermore, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0147] The units described as separate components may or may not be physically separate. A component shown as a unit can be one or more physical units; that is, it can be located in one place or distributed in multiple different locations. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0148] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0149] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, in essence, or the parts that contribute to the prior art, or all or part of the technical solutions, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0150] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An opportunity discovery processing system that integrates multidimensional data streams, characterized in that, include: A multidimensional data stream access module is used to acquire multidimensional data streams in real time, wherein the multidimensional data streams include multiple data dimensions that characterize the state of the business environment; The data discretization encoding module is used to perform discretization encoding operations for each data dimension in the multidimensional data stream to generate the corresponding binary encoded string; The unified context hash generation module is used to generate a unified context hash string based on multiple binary encoded strings corresponding to the multiple data dimensions; An opportunity discovery engine is used to perform matching queries in a historical context database based on the unified context hash string to identify marketing opportunities; The marketing solution generation module is used to generate marketing solutions in response to the marketing opportunities.

2. The system according to claim 1, characterized in that, The step of performing a discretization encoding operation for each data dimension in the multidimensional data stream to generate a corresponding binary encoded string includes: Determine the type of the data dimension; When the data dimension is of the first type, the numerical data is mapped to a discrete interval according to a preset quantization boundary set, and the first binary code string corresponding to the discrete interval is obtained. When the data dimension is of type two, the categorized data is mapped to a preset enumeration value according to a preset enumeration mapping relationship, and the second binary encoding string corresponding to the preset enumeration value is obtained.

3. The system according to claim 1, characterized in that, The process of generating a unified context hash string based on the multiple binary encoded strings corresponding to the multiple data dimensions includes: Determine the maximum encoding length among the plurality of binary encoded strings; Starting from the first and second positions of the plurality of binary encoded strings in a preset dimensional order, the bits of each binary encoded string at the current position are extracted iteratively in positional order. The multiple bits extracted in each iteration are concatenated according to the preset dimensional order to form an intermediate string fragment; The intermediate string fragments generated in all iterations corresponding to the maximum encoding length are combined in sequence to form the unified context hash string.

4. The system according to claim 1, characterized in that, The step of performing a matching query in the historical context database based on the unified context hash string to identify marketing opportunities includes: Use the unified context hash string as the query key to perform an exact match query in the historical context database; If the query finds a historical context record and the historical business performance indicator associated with the historical context record is higher than a first preset threshold, then the marketing opportunity is identified as a historical recurrence opportunity.

5. The system according to claim 4, characterized in that, After using the unified context hash string as the query key to perform an exact match query in the historical context database, the system further includes: Extract a prefix of a preset length from the unified context hash string to obtain the context prefix string; Using the context prefix string as a query condition, a prefix matching query is performed in the historical context database to retrieve historical context records with the same prefix; Calculate the statistical values ​​of the historical business performance indicators associated with the historical context records; If the statistical value is higher than the second preset threshold, the marketing opportunity will be classified as a pattern-similar opportunity.

6. The system according to claim 4, characterized in that, After using the unified context hash string as the query key to perform an exact match query in the historical context database, the system further includes: Obtain the historical unified context hash string within a preset time window prior to the current time point to form the current context sequence; Determine whether the current context sequence matches a predefined value transition pattern; If a match is found, the marketing opportunity will be classified as a state transition opportunity.

7. The system according to claim 1, characterized in that, The marketing solution generated in response to the marketing opportunity includes: Select the corresponding plan template from the plan template library based on the type of marketing opportunity; Parse the parameter placeholders contained in the scheme template; The marketing solution is generated by filling the parameter placeholders with real-time data from the multidimensional data stream.

8. A method for opportunity discovery processing that integrates multidimensional data streams, characterized in that, include: Real-time acquisition of multidimensional data streams, which include multiple data dimensions characterizing the state of the business environment; For each data dimension in the multidimensional data stream, a discretization encoding operation is performed to generate a corresponding binary encoded string; A unified context hash string is generated based on the multiple binary encoded strings corresponding to the multiple data dimensions; Based on the unified context hash string, a matching query is performed in the historical context database to identify marketing opportunities; In response to the aforementioned marketing opportunities, generate marketing solutions.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein the computer program is configured to execute the method of claim 8 at runtime.

10. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to run the computer program to perform the method of claim 8.

Images