Purchase behavior discrimination method and system based on vector retrieval

Abstract

The invention discloses a purchase behavior discrimination method and system based on vector retrieval, and the method comprises the steps: carrying out the vectorization processing of the current multi-dimensional behavior data of a user, and obtaining the current user behavior vector of the user; performing similarity calculation on pre-stored user behavior vectors and the obtained user behavior vectors, and determining user behavior vectors with high similarity; and judging the current purchase behavior of the user according to the label of the user behavior vector with high similarity. According to the invention, the judgment accuracy and timeliness of the purchase behavior of the user can be improved.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a method and system for determining purchase behavior based on vector retrieval. Background Technology

[0002] In the retail and e-commerce industries, accurate identification of user purchasing behavior is the core foundation for achieving precision marketing, optimizing inventory management, and improving user conversion efficiency. By identifying users' potential purchasing tendencies, determining their purchase intentions, or identifying signals indicating abstention from purchase, merchants can target product information and adjust service strategies accordingly, thereby reducing marketing costs and increasing user satisfaction. However, existing purchasing behavior identification technologies have the following key problems: 1. Low accuracy: Traditional technologies are mostly based on rule engines (such as "orders placed within 24 hours of adding to cart are considered confirmed purchases") or simple statistical models (such as behavior frequency statistics). They can only capture explicit user behavior characteristics and cannot effectively mine the purchase intentions behind implicit characteristics such as browsing time, page jump path, and product comparison frequency. This leads to a large bias in the judgment of complex behaviors (such as "repeatedly browsing and comparing but not immediately adding to cart").

[0003] 2. Insufficient retrieval efficiency: As the user base expands and the dimensions of behavioral data increase (such as multi-source data such as browsing, adding to cart, and payment by tens of millions of users every day), traditional query methods based on relational databases are unable to cope with the rapid matching needs of massive vector data. The retrieval response time often reaches the second level, which cannot meet the timeliness requirements of real-time marketing scenarios. Summary of the Invention

[0004] In view of this, the purpose of this invention is to propose a method and system for judging purchase behavior based on vector retrieval, which can improve the accuracy and timeliness of judging user purchase behavior.

[0005] To achieve the above objectives, this invention provides a method for determining purchase behavior based on vector retrieval, comprising: The user's current multi-dimensional behavior data is vectorized to obtain the user's current user behavior vector. The similarity between the pre-stored user behavior vector and the obtained user behavior vector is calculated to determine the user behavior vector with high similarity. The user's current purchasing behavior is determined based on the labels of user behavior vectors with high similarity.

[0006] Preferably, determining the user behavior vectors with high similarity specifically involves: The calculated similarity scores are sorted in descending order. The user behavior vectors that rank highest in similarity are identified as those with high similarity to the current user behavior vector of the user.

[0007] Preferably, determining the user's current purchasing behavior based on the labels of user behavior vectors with high similarity specifically includes: The user behavior vectors with high similarity are identified as similar user behavior vectors; Calculate the average similarity between similar user behavior vectors and the current user behavior vector of the user to obtain the average similarity. Calculate the ratio of the number of user behavior vectors labeled "purchase" to the number of user behavior vectors labeled "not purchase" in similar user behavior vectors to obtain the purchase tag percentage; The user's current purchasing behavior is determined based on the calculated average similarity and the proportion of purchase tags.

[0008] Preferably, determining the user's current purchasing behavior based on the calculated average similarity and the proportion of purchase tags specifically includes: If the calculated average similarity is greater than or equal to the set similarity threshold, and the proportion of purchased tags is greater than or equal to the set proportion threshold, then the user's current purchase behavior is determined to be "confirm purchase".

[0009] This invention also provides a purchase behavior discrimination system based on vector retrieval, comprising: The vector construction module is used to vectorize the user's current multi-dimensional behavior data to obtain the user's current behavior vector. The vector retrieval module is used to calculate the similarity between the pre-stored user behavior vectors and the obtained user behavior vectors, and to determine the user behavior vectors with high similarity. The behavior discrimination module is used to determine the user's current purchasing behavior based on the labels of user behavior vectors with high similarity.

[0010] The present invention also provides a computer device, the computer device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described vector retrieval-based purchase behavior discrimination method.

[0011] The present invention also provides a computer-readable storage medium storing a computer program that can be executed by at least one processor to cause the at least one processor to perform the steps of the above-described vector retrieval-based purchase behavior discrimination method.

[0012] In this invention, the user's current multi-dimensional behavior data is vectorized to obtain the user's current behavior vector. A similarity calculation is performed between the pre-stored user behavior vector and the obtained user behavior vector to determine the user behavior vectors with high similarity. Based on the tags of the user behavior vectors with high similarity, the user's current purchasing behavior is determined. Compared to traditional techniques that are mostly based on rule engines, the comparison results between the user behavior vectors obtained from multi-dimensional behavior data and the retrieved user behavior vectors with high similarity show higher accuracy in identifying complex behaviors. Compared to traditional query methods based on relational databases, the vector retrieval method of this invention is faster and more efficient. Attached Figure Description

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1 A flowchart of a purchase behavior discrimination method based on vector retrieval provided in an embodiment of the present invention; Figure 2 A structural block diagram of a purchase behavior discrimination system based on vector retrieval provided in an embodiment of the present invention; Figure 3a An internal structure block diagram of a vector construction module provided in an embodiment of the present invention; Figure 3b An internal structural block diagram of a vector retrieval module provided in an embodiment of the present invention; Figure 3c An internal structural block diagram of a behavior discrimination module provided in an embodiment of the present invention; Figure 3d An internal structural block diagram of a data acquisition module provided in an embodiment of the present invention; Figure 3e An internal structural block diagram of a data preprocessing module provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of a computer device hardware structure provided in an embodiment of the present invention. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to specific embodiments and accompanying drawings.

[0016] It should be noted that, unless otherwise defined, the technical or scientific terms used in the embodiments of this invention should have the ordinary meaning understood by one of ordinary skill in the art to which this disclosure pertains. The terms "first," "second," and similar terms used in this disclosure do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.

[0017] The technical solutions of the embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0018] like Figure 1 As shown in the figure, the process of a purchase behavior discrimination method based on vector retrieval proposed in this embodiment of the invention is as follows: Figure 1 As shown, it includes the following steps: Step S101: Collect the user's current multi-dimensional behavioral data; Specifically, multi-dimensional behavioral data of users can be obtained through e-commerce platform interfaces, including user browsing history, interaction records, preference tags, and other data. For example, it can be obtained that user A browsed the "laptop" category page 12 times in the past 7 days (average dwell time of 45 seconds), added 2 models to the cart (dwelling for 180 seconds and 240 seconds respectively), and did not place an order. In addition, redundant data can be filtered out from the collected data to obtain the effective data, and then the effective data can be stored in HDFS (Hadoop Distributed File System).

[0019] Step S102: Preprocess the collected multi-dimensional behavioral data of the user; In this step, the preprocessing of the collected multi-dimensional user behavior data includes: cleaning the collected data to remove redundant data, performing data standardization, and then handling outliers.

[0020] Step S103: Vectorize the user's current multi-dimensional behavior data to obtain the user's current behavior vector; In this step, the user's current multi-dimensional behavior data is vectorized: the core features of the collected data are extracted (e.g., browsing the laptop 12 times in the past 7 days, adding 2 items to the cart, and an average browsing-to-cart interval of 30 minutes), and then the features are encoded into multi-dimensional vectors (e.g., 128-dimensional vectors) using the Word2Vec (word-to-vector) model; then the obtained multi-dimensional vectors are normalized to obtain normalized vectors, which serve as the optimized user behavior vectors.

[0021] Step S104: Calculate the similarity between the pre-stored user behavior vector and the obtained user behavior vector, and determine the user behavior vector with high similarity. In this step, the pre-stored user behavior vectors in the historical vector library are called and their similarity is calculated with the user's current user behavior vector. The calculated similarity is sorted in descending order. The user behavior vectors with the highest similarity are determined as user behavior vectors with high similarity to the user's current user behavior vector.

[0022] Step S105: Determine the user's current purchasing behavior based on the labels of user behavior vectors with high similarity; In this step, the user behavior vectors with high similarity are identified as similar user behavior vectors; the average similarity between similar user behavior vectors and the user's current user behavior vector is calculated to obtain the average similarity; the ratio of the number of user behavior vectors labeled "purchase" to the number of user behavior vectors labeled "not purchase" in the similar user behavior vectors is calculated to obtain the purchase tag ratio; based on the calculated average similarity and purchase tag ratio, the user's current purchase behavior is determined.

[0023] Specifically, if the calculated average similarity is greater than or equal to a set similarity threshold, and the proportion of purchase tags is greater than or equal to a set proportion threshold, then the user's current purchase behavior is determined to be "confirm purchase"; otherwise, the user's current purchase behavior is determined to be "confirm not to purchase".

[0024] This invention takes into account that existing systems mostly use fixed models and retrieval parameters, and fail to establish a model iteration and parameter update mechanism based on incremental data to address the dynamic changes in user behavior preferences over time and in different scenarios (such as promotional seasons and holidays), resulting in a continuous decline in discrimination accuracy after long-term use; Based on this, in the technical solution of the present invention, for the purchase behavior of a user determined to "confirm purchase", if it is subsequently confirmed that the user has indeed performed the "purchase" behavior, the user behavior vector corresponding to the purchase behavior of the user determined to "confirm purchase" is labeled "purchase" and stored in the historical vector library; if it is subsequently confirmed that the user has not performed the "purchase" behavior, the user behavior vector corresponding to the purchase behavior of the user determined to "confirm purchase" is labeled "not purchase" and stored in the historical vector library; thus, an iterative and parameter update mechanism for the historical vector library is established to address the dynamic changes in user behavior preferences over time and in different scenarios (such as promotional seasons and holidays).

[0025] Based on the above-described vector-based purchase behavior discrimination method, this invention proposes a vector-based purchase behavior discrimination system, the internal structure of which is shown in the following diagram. Figure 2 As shown, it includes the following modules: vector construction module 203, vector retrieval module 204, and behavior discrimination module 205; Vector construction module 203 is used to vectorize the user's current multi-dimensional behavior data to obtain the user's current behavior vector; specifically, as shown in the example... Figure 3a As shown, the vector construction module 203 may include the following units: feature extraction unit 31, vector encoding unit 32, and vector optimization unit 33; Among them, the feature extraction unit 31 is used to extract the core features of the collected data; Vector encoding unit 32 is used to encode features into multidimensional vectors using the Word2Vec (Word Turning Vector) model; The vector optimization unit 33 normalizes the obtained multidimensional vectors to obtain normalized vectors, which are used as optimized user behavior vectors.

[0026] Vector retrieval module 204 is used to calculate the similarity between pre-stored user behavior vectors and user behavior vectors obtained by vector construction module 203, and to determine user behavior vectors with high similarity; specifically, such as Figure 3b As shown, the vector retrieval module 204 may include the following units: vector database unit 41, near nearest neighbor retrieval unit 42, and retrieval efficiency optimization unit 43; Among them, the vector database unit 41 is used to call the user behavior vectors pre-stored in the historical vector library; The near nearest neighbor retrieval unit 42 is used to calculate the similarity between each user behavior vector pre-stored in the historical vector database and the user's current user behavior vector; the calculated similarity is sorted in descending order; the user behavior vectors with the highest similarity are determined as user behavior vectors with high similarity to the user's current user behavior vector; for example, the FAISS (Facebook AI Similarity Search) algorithm is used to retrieve the Top 50 similar vectors, of which 38 are tagged "purchase" and 12 are "not purchased", with an average similarity of 0.85; The retrieval efficiency optimization unit 43 is used to control the retrieval time to 80ms through IVF (Inverted File Index).

[0027] The behavior determination module 205 is used to determine the user's current purchasing behavior based on the labels of user behavior vectors with high similarity determined by the vector retrieval module 204; specifically, as follows: Figure 3c As shown, the behavior discrimination module 205 may include the following units: feature matching unit 51, behavior category division unit 52, and discrimination threshold adjustment unit 53; The feature matching unit 51 is used to take the user behavior vectors with high similarity as similar user behavior vectors; calculate the average similarity between similar user behavior vectors and the current user behavior vector of the user to obtain the average similarity; calculate the ratio of the number of user behavior vectors with the tag "purchase" to the number of user behavior vectors with the tag "not purchase" in the similar user behavior vectors to obtain the purchase tag ratio; for example, the calculated purchase tag ratio is 76% (38 / 50) and the average similarity is 0.85. The behavior category segmentation unit 52 is used to determine the user's current purchase behavior based on the calculated average similarity and the proportion of purchase tags; for example, it determines the user as "confirmed to purchase" based on the rule (similarity ≥ T1 and tag proportion ≥ T2); Where T1 is the similarity threshold, which is used to measure the degree of matching between the current user behavior vector and the historical similar user behavior vector, and its value ranges from [0,1]. T2 is the purchase tag percentage threshold, used to measure the percentage of purchase behavior among similar historical user behaviors, with a value range of [0%, 100%].

[0028] Initially, the course can be manually set by technical personnel according to e-commerce business scenarios (such as product category characteristics and marketing cycles), for example, the initial default value (T1=0.8, T2=70%).

[0029] The similarity threshold T1 is related to "average similarity = total similarity / number of similar vectors": T1 is the passing line for "average similarity" and is used to filter out historical similar vector groups that meet the matching degree with the current user behavior (for example, when T1=0.8, only similar vectors with average similarity ≥0.8 are retained, and invalid data with low matching degree are filtered out). The purchase tag percentage threshold T2 is related to "purchase tag percentage = number of purchase tag vectors / total number of similar vectors": T2 is the passing line for "purchase tag percentage", used to filter out groups whose purchase behavior percentage in historical similar vectors meets the standard (for example, when T2=70%, only similar vectors with purchase tag percentage ≥70% are retained to avoid interference from low purchase tendency groups in the judgment). The system must simultaneously meet the criteria of "average similarity ≥ T1" and "purchase tag percentage ≥ T2" to determine the corresponding behavior category (such as "confirm purchase"). If either threshold is not met, the category cannot be determined to ensure accuracy.

[0030] The discrimination threshold adjustment unit 53 is used to determine whether the current purchase behavior discrimination threshold needs to be adjusted based on the recent "confirm purchase" user conversion rate (85%), that is, the ratio of users who were judged as "confirm purchase" by the behavior category division unit 52 and who actually carried out the purchase behavior.

[0031] Specifically, the aforementioned purchase behavior discrimination threshold is the core criterion for accurately determining the category of user purchase behavior. It is not a single numerical threshold, but a binary combination threshold composed of a similarity threshold (denoted as T1) and a purchase tag proportion threshold (denoted as T2), denoted as (T1, T2).

[0032] The core criterion for adjusting the purchase behavior discrimination threshold is the actual conversion rate of users recently identified as belonging to the corresponding behavior category (e.g., "Confirm Purchase"). For example, if the actual conversion rate of users in the "Confirm Purchase" category deviates significantly from the system's expectations (e.g., expected conversion rate of 80%, actual rate of only 60%), then the purchase behavior discrimination threshold needs to be adjusted; if the actual conversion rate is close to the expectations (e.g., 85% in the example meets the expectations), then no adjustment is needed.

[0033] The adjustment of the purchase behavior discrimination threshold is executed by the discrimination threshold adjustment unit in the behavior discrimination module. For example, if the actual conversion rate is lower than expected, the threshold can be increased (e.g., adjusting "similarity ≥ 0.8 and tag percentage ≥ 70%" to "similarity ≥ 0.85 and tag percentage ≥ 75%) to filter more precise target users; if the actual conversion rate is higher than expected but the number of users covered is too small, the threshold can be appropriately decreased to expand the target user range.

[0034] In one implementation, the following similarity threshold adjustment formula and purchase tag proportion threshold adjustment formula can be used to dynamically adjust the purchase behavior discrimination threshold: Similarity threshold adjustment formula: ; Formula for adjusting the purchase tag percentage threshold: ; Among them, T1 new T1 is the adjusted similarity threshold. old The similarity threshold before adjustment; T2 new T2 is the adjusted threshold for the percentage of purchase tags. old The threshold for the percentage of purchase tags before adjustment; This represents the actual conversion rate of users who recently "confirmed their purchase". The system's preset expected conversion rate (e.g., 80%).

[0035] Furthermore, the purchase behavior discrimination system based on vector retrieval proposed in this embodiment of the invention also includes: a data acquisition module 201, a data preprocessing module 202, a dynamic update module 207, and a result output module 206; Among them, the data acquisition module 201 is used to collect multi-dimensional behavioral data of the user; specifically, such as Figure 3d The data acquisition module 201 shown specifically includes the following units: multi-source data acquisition unit 11, real-time acquisition control unit 12, and data storage unit 13; Among them, the multi-source data acquisition unit 11 is used to obtain multi-dimensional behavioral data of users through the e-commerce platform interface, including data such as users' browsing trajectory, interaction records, and preference tags; The real-time acquisition control unit 12 is used to set the acquisition frequency, for example, to 2 minutes / time, thereby filtering redundant data in the acquired data, such as filtering out 3 redundant data generated by the user repeatedly refreshing the page, and obtaining valid data; Data storage unit 13 is used to store valid data to the HDFS distributed database.

[0036] The aforementioned data preprocessing module 202 is used to preprocess the collected multi-dimensional behavioral data of the user; specifically, such as... Figure 3e The data preprocessing module 202 shown includes the following units: data cleaning unit 21, data standardization unit 22, and outlier handling unit 23; The data cleaning unit 21 is used to clean the data collected by the data acquisition module 201 and remove redundant data. For example, it can remove 3 redundant data and fill in the missing data of 1 browsing time of the user (filled with an average of 45 seconds). The data standardization unit 22 is used to perform data standardization processing on the data cleaned by the data cleaning unit 21; for example, converting "12 times of browsing" into a ratio of 2.0 to the user's historical average browsing time (6 times), and then normalizing it to the [0,1] interval to get 0.8, and normalizing "2 items added to cart" to get 0.6; After determining that the user's behavior data is not abnormal according to the 3σ principle, the outlier processing unit 23 outputs standardized data.

[0037] The aforementioned result output module 206 is used to push the judgment result of the behavior judgment module 205 to the merchant; specifically, the output module 206 includes the following units: merchant-side output unit 61, data visualization unit 62, and early warning unit 63; Among them, the merchant-side output unit 61 is used to push the user's "confirm purchase" judgment result to the merchant; Data visualization unit 62 generates a heatmap of user behavior (showing that 7-9 PM is the peak browsing time each day); Warning unit 63 triggered a shipping preparation warning.

[0038] The aforementioned dynamic update module 207 is used to update the historical vector database according to the judgment result; specifically, the dynamic update module 207 includes the following units: incremental analysis unit 71, vector model iteration unit 72, and retrieval parameter update unit 73; Among them, the incremental analysis unit 71 incorporates user behavior data into the daily incremental data; Vector model iteration unit 72 will be fine-tuned with incremental data in two weeks; The retrieval parameter update unit 73 synchronously updates the FAISS index of the historical vector library.

[0039] In the technical solution of this invention, the user's current multi-dimensional behavior data is vectorized to obtain the user's current behavior vector; the pre-stored user behavior vector and the obtained user behavior vector are compared to calculate the similarity, and the user behavior vector with high similarity is determined; the user's current purchase behavior is determined based on the label of the user behavior vector with high similarity.

[0040] The accuracy of the technical solution of this invention is significantly improved: by vectorizing the multi-dimensional behavioral characteristics of users (such as browsing-add to cart time interval), and combining the similarity matching of the FAISS algorithm, the accuracy of the judgment of "potential purchase" and "confirmed purchase" behaviors is improved to more than 90%, which is 30% higher than that of traditional rule engines.

[0041] The retrieval efficiency of the technical solution of this invention is greatly optimized: by adopting near nearest neighbor retrieval and index optimization technology, millisecond-level retrieval response is achieved under massive data, which meets the timeliness requirements of real-time marketing on e-commerce platforms.

[0042] The technical solution of this invention has strong dynamic adaptability: the model and parameters are iteratively optimized through the dynamic update module, which can adapt to the trend changes of user behavior in real time (such as promotional seasons and new product launches), and the discrimination accuracy decay rate is less than 5% under long-term use.

[0043] The technical solution of this invention has high application value: multi-source data collection and visualization output can help merchants formulate precise marketing strategies, such as pushing coupons to "potential buyers" and preparing goods in advance for "confirmed buyers", helping merchants improve conversion rate and user satisfaction, but cannot adapt to the dynamic evolution of user behavior.

[0044] Figure 4 This illustration schematically shows the hardware architecture of a computer device 1300 for a purchase behavior discrimination method based on vector retrieval according to an embodiment of this application. In this embodiment, the computer device 1300 is a device capable of automatically performing numerical calculations and / or information processing according to pre-set or stored instructions. For example, it may be a smartphone, tablet computer, laptop computer, desktop computer, rack server, blade server, tower server, or cabinet server (including independent servers or server clusters composed of multiple servers), etc. Figure 4 As shown, the computer device 1300 includes, but is not limited to, at least: a memory 1310, a processor 1320, and a network interface 1330 that can communicate with each other via a system bus. Wherein: The memory 1310 includes at least one type of computer-readable storage medium, including flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, optical disk, etc. In some embodiments, the memory 1310 may be an internal storage module of the computer device 1300, such as the hard disk or memory of the computer device 1300. In other embodiments, the memory 1310 may also be an external storage device of the computer device 1300, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. Of course, the memory 1310 may also include both the internal storage module and the external storage device of the computer device 1300. In this embodiment, the memory 1310 is typically used to store the operating system and various application software installed on the computer device 1300, such as the program code for a purchase behavior discrimination method based on vector retrieval. In addition, the memory 1310 can also be used to temporarily store various types of data that have been output or will be output.

[0045] In some embodiments, processor 1320 may be a central processing unit (CPU), controller, microcontroller, microprocessor, or other data processing chip. Processor 1320 is typically used to control the overall operation of computer device 1300, such as performing control and processing related to data interaction or communication with computer device 1300. In this embodiment, processor 1320 is used to run program code stored in memory 1310 or process data.

[0046] Network interface 1330 may include a wireless network interface or a wired network interface, which is typically used to establish a communication link between computer device 1300 and other computer devices. For example, network interface 1330 is used to connect computer device 1300 to an external terminal via a network, establishing a data transmission channel and communication link between computer device 1300 and the external terminal. The network may be an intranet, the Internet, Global System for Mobile Communication (GSM), Wideband Code Division Multiple Access (WCDMA), 4G network, 5G network, Bluetooth, Wi-Fi, or other wireless or wired networks.

[0047] It should be pointed out that, Figure 4 Only a computer device with components 1310-1330 is shown; however, it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0048] In this embodiment, the purchase behavior discrimination method based on vector retrieval stored in memory 1310 can be further divided into one or more program modules and executed by one or more processors (processor 1320 in this embodiment) to complete the embodiment of this application.

[0049] The computer-readable medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0050] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of this disclosure (including the claims) is limited to these examples; within the framework of this invention, the technical features of the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other variations of the different aspects of the invention as described above, which are not provided in the details for the sake of brevity.

[0051] Additionally, to simplify the description and discussion, and to avoid obscuring the invention, the well-known power / ground connections to integrated circuit (IC) chips and other components may or may not be shown in the provided drawings. Furthermore, the apparatus may be shown in block diagram form to avoid obscuring the invention, and this also takes into account the fact that the details of implementation of these block diagram apparatuses are highly dependent on the platform on which the invention will be implemented (i.e., these details should be fully understood by those skilled in the art). While specific details (e.g., circuits) are set forth to describe exemplary embodiments of the invention, it will be apparent to those skilled in the art that the invention may be practiced without these specific details or with variations thereof. Therefore, these descriptions should be considered illustrative rather than restrictive.

[0052] Although the invention has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory architectures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed.

[0053] The embodiments of this invention are intended to cover all such substitutions, modifications, and variations falling within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for determining purchase behavior based on vector retrieval, characterized in that, include: The user's current multi-dimensional behavior data is vectorized to obtain the user's current user behavior vector. The similarity between the pre-stored user behavior vector and the obtained user behavior vector is calculated to determine the user behavior vector with high similarity. The user's current purchasing behavior is determined based on the labels of user behavior vectors with high similarity.

2. The method according to claim 1, characterized in that, The process of determining user behavior vectors with high similarity is as follows: The calculated similarity scores are sorted in descending order. The user behavior vectors that rank highest in similarity are identified as those with high similarity to the current user behavior vector of the user.

3. The method according to claim 2, characterized in that, The step of determining the user's current purchasing behavior based on the labels of highly similar user behavior vectors specifically includes: The user behavior vectors with high similarity are identified as similar user behavior vectors; Calculate the average similarity between similar user behavior vectors and the current user behavior vector of the user to obtain the average similarity. Calculate the ratio of the number of user behavior vectors labeled "purchase" to the number of user behavior vectors labeled "not purchase" in similar user behavior vectors to obtain the purchase tag percentage; The user's current purchasing behavior is determined based on the calculated average similarity and the proportion of purchase tags.

4. The method according to claim 3, characterized in that, The step of determining the user's current purchasing behavior based on the calculated average similarity and the proportion of purchase tags specifically includes: If the calculated average similarity is greater than or equal to the set similarity threshold, and the proportion of purchased tags is greater than or equal to the set proportion threshold, then the user's current purchase behavior is determined to be "confirm purchase".

5. The method according to claim 1, characterized in that, The multi-dimensional behavioral data specifically includes: browsing trajectory, interaction records, and preference tag data.

6. A purchase behavior discrimination system based on vector retrieval, characterized in that, include: The vector construction module is used to vectorize the user's current multi-dimensional behavior data to obtain the user's current behavior vector. The vector retrieval module is used to calculate the similarity between the pre-stored user behavior vectors and the obtained user behavior vectors, and to determine the user behavior vectors with high similarity. The behavior discrimination module is used to determine the user's current purchasing behavior based on the labels of user behavior vectors with high similarity.

7. The system according to claim 6, characterized in that, The vector retrieval module is specifically used to sort the calculated similarities in descending order; and to determine the user behavior vectors with the highest similarity ranking as user behavior vectors with high similarity to the current user behavior vector.

8. The system according to claim 6, characterized in that, The behavior discrimination module is specifically used to take the user behavior vectors with high similarity as similar user behavior vectors; calculate the average similarity between the similar user behavior vectors and the user's current user behavior vector to obtain the average similarity; Calculate the ratio of the number of user behavior vectors labeled "purchase" to the number of user behavior vectors labeled "not purchase" in the similar user behavior vectors to obtain the purchase tag ratio; determine the user's current purchase behavior based on the calculated average similarity and purchase tag ratio.

9. A computer device, the computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the purchase behavior discrimination method based on vector retrieval as described in any one of claims 1 to 4.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that can be executed by at least one processor to cause the at least one processor to perform the steps of the purchase behavior discrimination method based on vector retrieval as described in any one of claims 1 to 4.

Images