Object classification method and device and processor

Abstract

The invention discloses an object classification method and device and a processor. The method comprises the following steps: acquiring historical transaction information, preference information and account intra-line information corresponding to a target account in a target account set; based on the historical transaction information, the preference information and the account intra-line information, an interest sequence of the target account is constructed, and the interest sequence is used for representing the change state of the interest content of the target account in the time dimension; public interest subsequences are determined from the interest sequences corresponding to the target accounts in the target account set, and the public interest subsequences are the same interest subsequences in the interest sequences of different target accounts; and based on the attribute information of the public interest subsequences, determining interest similarity among different target accounts, and based on the interest similarity, classifying the target accounts to obtain a classification result. Through the object classification method and device, the technical problem of low object classification accuracy in related technologies is solved.

Description

Technical Field

[0001] The present application relates to the field of financial technology, and more specifically, to a method, device, and processor for classifying objects. Background Art

[0002] Currently, banking marketing recommendation systems play a vital role in customer marketing. On the one hand, through big data analysis, recommendation systems can identify customer preferences, helping banks improve marketing efficiency. On the other hand, recommendation systems help customers better understand banking products and enhance their banking experience. Recommendation systems involve multiple steps, including information extraction, customer modeling, content optimization, and user preference prediction, and are the nexus between banking marketing and customer experience. Customer modeling and preference prediction are two core issues in recommendation systems. During the modeling process, recommendation systems compare the behavioral characteristics of different customers with those of target customers to determine the model structure and parameters. During the preference prediction phase, the system exploits patterns in the established model to predict target customers' interest in future or unknown products. In banking marketing, recommendation systems can predict customer interest in unknown banking products. Customer-centric recommendation systems predict product preferences based on customer behavior. Using personalized recommendation mechanisms, they facilitate matching bank customers with bank products, improving marketing efficiency and customer satisfaction.

[0003] In related technologies, recommendation systems are primarily categorized into collaborative filtering and content filtering. Collaborative filtering algorithms predict a user's interest in unconsumed items based on similarities between users or items, while content filtering predicts user interest based on the attributes of the items themselves. Both types of recommendation algorithms aim for prediction. Collaborative filtering algorithms have relatively low prediction accuracy. Specifically, predicting user interests based on similarities between users can lead to biased recommendation results; predicting user interests based on similarities between items can lead to duplicated recommendation results; and predicting based on a user's historical transaction data can lead to cold start issues. Consequently, the technical issue of low object classification accuracy persists.

[0004] Currently, no effective solution has been proposed to the technical problem of low classification accuracy of objects in related technologies. Summary of the Invention

[0005] The main purpose of this application is to provide a method, device and processor for classifying objects to solve the technical problem of low object classification accuracy in related technologies.

[0006] To achieve the above-mentioned objectives, according to one aspect of the present application, a method for classifying objects is provided. The method comprises: obtaining historical transaction information, preference information, and account line information corresponding to a target account in a target account set, wherein the historical transaction information is used to represent the target account's historical transaction behavior in a service system, and the account line information is used to represent the attributes of the target account's registration in the service system; constructing an interest sequence for the target account based on the historical transaction information, preference information, and account line information, wherein the interest sequence is used to represent the changing state of the target account's interest content over time; determining common interest subsequences from the interest sequences corresponding to the target account in the target account set, wherein the common interest subsequences are the same interest subsequences in the interest sequences of different target accounts; determining interest similarities between different target accounts based on the attribute information of the common interest subsequences, and classifying the target accounts based on the interest similarities to obtain classification results, wherein the interest similarity is used to represent the degree of similarity between different target accounts for the same interest content.

[0007] Optionally, based on historical transaction information, preference information and account line information, an interest sequence of the target account is constructed, including: based on historical transaction information, preference information and account line information, constructing interest points for the target account, wherein the interest points are used to represent the target account's evaluation behavior on the service item at the timestamp; sorting the interest points in order of the timestamps to obtain an interest sequence.

[0008] Optionally, a common interest subsequence is determined from the interest sequence corresponding to the target account in the target account set, including: obtaining the interest subsequence in the interest sequence; matching the interest subsequences corresponding to different target accounts in the target account set to obtain rating difference data, wherein the rating difference data is used to indicate the degree of difference in ratings of different target accounts for the same service item; and determining the common interest subsequence based on the rating difference data.

[0009] Optionally, the interest sub-sequences corresponding to different target accounts in the target account set are matched to obtain matching results, including: matching the interest sub-sequences corresponding to the evaluation behaviors performed by different target accounts on the same service item, and determining the score difference data corresponding to the interest sub-sequences corresponding to the same service item of different target accounts.

[0010] Optionally, determining the common interest subsequence based on the score difference data includes: determining the interest subsequence corresponding to the score difference data being less than or equal to a score difference threshold as the common interest subsequence.

[0011] Optionally, the attribute information includes target length information and quantity information of the longest common interest subsequence, wherein, before determining the interest similarity between different target accounts based on the attribute information of the common interest subsequence, and classifying the target accounts based on the interest similarity, the method also includes: obtaining the length information of the common interest subsequence; determining the longest common interest subsequence from the common interest subsequence based on the length information, and obtaining the target length information of the longest common interest subsequence; obtaining the quantity information of the common interest subsequences between two target accounts in the target account set; the interest similarity includes a first interest similarity and a second interest similarity, wherein, determining the interest similarity between different target accounts based on the attribute information of the common interest subsequence includes: determining the first interest similarity based on the target length information; determining the second interest similarity based on the quantity information; and determining the interest similarity based on the first interest similarity and the second interest similarity.

[0012] Optionally, the target accounts are classified based on interest similarity, including: integrating the interest similarity into the Pearson coefficient to obtain target interest similarity; and classifying the target accounts based on the target interest similarity.

[0013] Optionally, the method further includes: formulating corresponding marketing strategies for target accounts of different categories, and recommending service items to the target accounts according to the marketing strategies.

[0014] To achieve the above-mentioned object, according to another aspect of the present application, a device for classifying objects is provided. The device comprises: an acquisition unit configured to acquire historical transaction information, preference information, and account line information corresponding to a target account in a target account set, wherein the historical transaction information represents the target account's historical transaction behavior in a service system, and the account line information represents the attributes of the target account registered in the service system; a construction unit configured to construct an interest sequence of the target account based on the historical transaction information, preference information, and account line information, wherein the interest sequence represents the changing state of the target account's interest content over time; a determination unit configured to determine a common interest subsequence from the interest sequences corresponding to the target account in the target account set, wherein the common interest subsequence is the same interest subsequence in the interest sequences of different target accounts; and a classification unit configured to determine interest similarity between different target accounts based on the attribute information of the common interest subsequence, and classify the target accounts based on the interest similarity to obtain a classification result, wherein the interest similarity represents the degree of similarity between different target accounts for the same interest content.

[0015] According to another aspect of an embodiment of the present application, a processor is further provided, which is configured to run a program, wherein when the program is run by the processor, the object classification method according to the embodiment of the present application is executed.

[0016] According to another aspect of the embodiments of the present application, an electronic device is further provided, including: a memory storing an executable program; and a processor for running the program, wherein the object classification method of each embodiment of the present application is executed when the program is running.

[0017] According to another aspect of the embodiments of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the object classification method in the embodiments of the present application.

[0018] According to another aspect of an embodiment of the present application, a computer program product is further provided. The computer program product includes a computer program, wherein when the computer program is executed by a processor, the object classification method in the embodiment of the present application is implemented.

[0019] In an embodiment of the present application, if it is necessary to categorize target accounts in a target account set to achieve precision marketing, historical transaction information, preference information, and account-line information corresponding to each target account can be obtained. Based on the historical transaction information, preference information, and account-line information, the changes in the target account's preferred content over different time periods can be analyzed to obtain interest sequences for each target account. The interest sequences of each target account can be matched to determine whether the preferred content of different target accounts is the same. Based on the same content, a common interest subsequence can be determined. Based on the attribute information of the common interest subsequence, the degree of interest similarity between the preferred content of different target accounts can be determined. Based on the interest similarity, the target accounts can be categorized. That is, target accounts with similar content and similar preferences can be grouped into the same category to obtain a classification result, thereby providing targeted marketing for target accounts in different categories. In this embodiment, by fusing multi-dimensional data to construct interest sequences, determine common interest subsequences, calculate comprehensive interest similarities and perform classification based on this, not only the accuracy of classification is improved, but also a more effective precision marketing tool is provided for banks, which can better meet user needs and improve user satisfaction, solve the technical problem of low classification accuracy of objects in related technologies, and achieve the technical effect of improving the classification accuracy of objects. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings, which constitute part of this application, are intended to provide a further understanding of this application. The exemplary embodiments and descriptions of this application are intended to explain this application and do not constitute an improper limitation on this application. In the accompanying drawings:

[0021] Figure 1 A hardware structure block diagram of a computer terminal for implementing an object classification method is shown;

[0022] Figure 2 is a flowchart of a method for classifying objects according to an embodiment of the present application;

[0023] Figure 3 This is a flowchart of a method for applying interest sequence-based collaborative filtering in bank marketing activities according to an embodiment of the present application;

[0024] Figure 4 is a schematic diagram of an object classification device provided according to an embodiment of the present application;

[0025] Figure 5 This is a structural block diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0026] In order to enable those skilled in the art to better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0027] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in a sequence other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0028] It should be noted that the collected information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for display, data for analysis, etc.) involved in this application are information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data comply with relevant laws, regulations and standards, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation portals for users to choose to authorize or refuse. For example, an interface is set up between this system and relevant users or institutions to provide users with corresponding operation portals for users to choose to agree or refuse the automated decision-making results; if the user chooses to refuse, the expert decision-making process will be entered.

[0029] Example 1

[0030] According to an embodiment of the present application, a method embodiment for classifying objects is also provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0031] The method embodiment provided in the first embodiment of the present application can be executed in a mobile terminal, a computer terminal or a similar computing device. Figure 1 FIG. 1 shows a hardware structure block diagram of a computer terminal (or mobile device) for implementing an object classification method. Figure 1 As shown, the computer terminal 10 (or mobile device) may include one or more (illustrated as 102a, 102b, ..., 102n in the figure) processors 102 (the processor 102 may include but is not limited to a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the BUS bus), a network interface, a power supply and / or a camera. It will be understood by those skilled in the art that Figure 1 The structure shown is only for illustration and does not limit the structure of the above electronic device. Figure 1 More or fewer components than shown, or with Figure 1 Different configurations shown.

[0032] It should be noted that the one or more processors 102 and / or other data processing circuits described above may generally be referred to herein as "data processing circuitry". The data processing circuitry may be embodied in whole or in part as software, hardware, firmware, or any other combination thereof. In addition, the data processing circuitry may be a single independent processing module, or may be incorporated in whole or in part into any of the other components of the computer terminal 10 (or mobile device). As described in the embodiments of the present application, the data processing circuitry serves as a processor control (e.g., selection of a variable resistor terminal path connected to an interface).

[0033] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the object classification method in the embodiment of the present application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, that is, implementing the above-mentioned object classification method. The memory 104 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include a memory remotely located relative to the processor 102, and these remote memories may be connected to the computer terminal 10 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0034] The transmission device 106 is used to receive or send data via a network. A specific example of the aforementioned network may include a wireless network provided by the communications provider of the computer terminal 10. In one embodiment, the transmission device 106 includes a network interface controller (NIC), which can be connected to other network devices via a base station to enable communication with the Internet. In another embodiment, the transmission device 106 may be a radio frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0035] The display may be, for example, a touch screen liquid crystal display (LCD), which enables a user to interact with a user interface of the computer terminal 10 (or mobile device).

[0036] Under the above operating environment, this application provides Figure 2 The classification method of the object shown. Figure 2 is a flow chart of a method for classifying objects according to an embodiment of the present application. Figure 2 As shown, the method may include the following steps:

[0037] Step S201 , obtaining historical transaction information, preference information, and account bank information corresponding to a target account in a target account set.

[0038] In step S201 of the embodiment of the present application, the target account set refers to a group of user accounts selected in the bank or service system. The above accounts usually have certain common characteristics or are within the target range of a specific marketing activity. The determination of the target account set is based on the marketing needs or strategies of the bank. For example, it may be users whose activity has decreased over a period of time, or a potential user group that the bank hopes to focus on promoting a certain service. The definition of the target account set is the starting point of the entire recommendation system process. It determines which users will be included in the subsequent data analysis and strategy formulation process. The target account is a single user account in the target account set. In the embodiment of the present application, the target account is the subject of analysis and classification by the bank or service system. By analyzing the detailed information of each target account, the bank can identify the specific needs and interests of the user, thereby conducting more accurate classification and marketing. That is, the target account can be an account registered by the bank's customers.

[0039] Optionally, historical transaction information is used to represent the target account's historical transaction behavior within the service system. This information covers all transaction records of the target account within the bank or service system, including but not limited to purchases, deposits and withdrawals, transfers, and browsing history. This information not only reveals the user's transaction habits and preferences, but also reflects characteristics such as their spending power and risk appetite. Obtaining historical transaction information is crucial for constructing user behavior patterns and interest sequences, and serves as the foundation for subsequent analysis of evolving user interests. Preference information, also known as customer social preference information, can include user behavior data outside of the bank or service system. This information complements transaction information within the bank system, providing a more comprehensive user profile. By analyzing preference information, banks can not only understand users' interests and hobbies, but also gain insights into their social networks and influence, which is crucial for formulating marketing strategies. Intra-account information represents the attributes of the target account registered within the service system. This information contains attribute data of the target account within the bank system. This information is crucial for classifying users and assessing their credit and asset status within the bank system.

[0040] In this embodiment, if the target accounts in the target account set need to be classified for targeted marketing, the historical transaction information, preference information, and account bank information corresponding to each target account in the target account set can be obtained.

[0041] Optionally, the service system (e.g., a bank) will collect relevant information about the target account from multiple sources, including historical transaction information, preference information, and account bank information. This step is the core of the data-driven marketing strategy, ensuring the accuracy of subsequent analysis and classification. Through comprehensive information collection, banks can build a comprehensive user profile based on transactions, preferences, and account attributes, providing data support for subsequent interest sequence construction, user similarity calculation, and classification. By integrating historical transaction information, preference information, and account bank information, banks can gain a deeper understanding of user needs and interest changes, thereby designing marketing strategies that are closer to users' actual needs and improving the conversion rate of marketing activities and user satisfaction.

[0042] Step S202: construct an interest sequence of the target account based on historical transaction information, preference information, and account bank information.

[0043] In step S202 of the embodiment of the present application, the interest sequence is used to represent the changing state of the interest content of the target account in the time dimension, and can also be called a user interest point set.

[0044] In this embodiment, after obtaining historical transaction information, preference information and account bank information corresponding to the target account in the target account set, an interest sequence of the target account can be constructed based on the above historical transaction information, preference information and account bank information.

[0045] Optionally, an interest sequence represents the target account's evolving interests over time, i.e., a collection of user interests. Constructing an interest sequence transforms multidimensional information into a time-series representation of user behavior and interest patterns. This step is crucial for capturing and quantifying user dynamics.

[0046] In the embodiment of the present application, the construction of the interest sequence is mainly based on the following aspects: historical transaction information, the transaction behavior of users in the bank's service system; preference information, the user's behavior data on external platforms such as social media, shopping platforms, and news websites; account bank information, including the personal information provided by the user when registering with the bank and the static attributes of the account.

[0047] Optionally, during the construction of interest sequences, the collected multi-source information can be integrated and processed to arrange the user's points of interest in chronological order to form a sequence. Each point of interest contains key information such as the user, item (or product), rating (or review), and timestamp. The rating can be the user's actual transaction record, evaluation level, or a rating simulated by preference information.

[0048] Optionally, the above process may involve multiple data preprocessing steps, such as data cleaning, outlier detection, score standardization, and time series data generation. For example, user behavior data on different platforms may differ in format, timestamp, and score scale, requiring unified formatting to ensure that the constructed interest series accurately reflects the dynamic changes in user interests.

[0049] In an embodiment of the present application, by arranging user behaviors in a time series, the bank can capture the trend of user interests changing over time, which is extremely critical for understanding the evolution of user needs and formulating adaptive marketing strategies. The interest sequence provides a continuous representation of user interests, which helps to identify patterns that are similar to other users' interests, thereby providing richer input for personalized recommendation algorithms and improving the accuracy of recommendations. Based on the interest sequence, the bank can segment users and identify user groups that are interested in specific marketing activities or products, so as to design more targeted marketing strategies and improve the participation and conversion rate of marketing activities. For new users or new products, the construction of the interest sequence can use the user's historical behavior and preference information to quickly establish a user portrait, provide the recommendation system with an initial similarity calculation basis, and alleviate the cold start effect. The changing trend of the interest sequence helps banks track the development of user interests, thereby better managing the user's service experience throughout their life cycle and providing timely product recommendations and customer care.

[0050] Step S203: determining a common interest subsequence from the interest sequences corresponding to the target accounts in the target account set.

[0051] In step S203 of the embodiment of the present application, the common interest subsequence is the same interest subsequence in the interest sequences of different target accounts, and can also be called a common subsequence.

[0052] In this embodiment, after constructing the interest sequence of the target account based on historical transaction information, preference information, and account line information, a common interest subsequence may be determined from the interest sequence.

[0053] Optionally, the above embodiment is a key step in calculating user similarity, the primary goal of which is to identify common interest subsequences from the target account's interest sequence—i.e., the identical or similar portions of interest sequences across different users. The key to this step lies in understanding and applying the concept of common interest subsequences, as well as how to effectively identify the longest common interest subsequence and all common interest subsequences from multiple interest sequences, thereby quantifying interest similarity between users.

[0054] Alternatively, a common interest subsequence refers to one or more identical sequences of interest points within the interest sequences of two or more target accounts. An interest point consists of a user, item, rating, and timestamp, and can reflect a user's level of interest in a specific item at a specific time. In banking marketing scenarios, common interest subsequences can reveal shared transaction or browsing behavior among users, as well as preferences for the same or similar services.

[0055] Optionally, the process of identifying common interest subsequences involves comparing and matching interest sequences of different target accounts to find common patterns between them.

[0056] Optionally, since different users may have different rating scales for the same service item, it is necessary to set a rating difference threshold to determine whether two ratings can be considered the same. Items (content of interest) with a rating difference less than the above-mentioned rating difference threshold will be regarded as items of the same interest to the users. From each interest sequence, all possible subsequences are generated and matched with the interest sequences of other users to find the longest common subsequence and all common subsequences. Based on the length of the longest common interest subsequence found and the number of all common interest subsequences, the interest similarity between users is calculated. This step takes into account the depth (length of the longest common subsequence) and breadth (number of common subsequences) of interest, thereby comprehensively evaluating the similarity of user interests.

[0057] In an embodiment of the present application, by identifying common interest subsequences and finding common interest patterns among users, banks can more accurately predict users' potential interest in service items, thereby improving the accuracy of the recommendation algorithm. Based on the similarity of common interest subsequences, users with similar interests are grouped into the same group. Banks can customize targeted marketing strategies for each user group and improve the conversion rate of marketing activities. For new users or new products, similar common interest subsequences can be quickly identified by comparison with existing user interest sequences, thereby recommending service items that may be of interest to new users and alleviating the "cold start" problem. By accurately identifying user interests, banks can optimize customer experience, provide services that meet user expectations, and enhance user satisfaction and loyalty.

[0058] Step S204 : determining the interest similarity between different target accounts based on the attribute information of the common interest subsequence, and classifying the target accounts based on the interest similarity to obtain a classification result.

[0059] In step S204 of the embodiment of the present application, interest similarity can be used to indicate the degree of similarity between different target accounts for the same interest content. Attribute information can be attributes such as the length and number of common interest subsequences, which is only for example and not specific.

[0060] In this embodiment, after determining the common interest subsequence from the interest sequence corresponding to the target account in the target account set, the interest similarity between different target accounts can be determined based on the attribute information of the common interest subsequence, so that the interest similarity can be used to classify target accounts with high similarity in interest content into the same category to obtain classification results.

[0061] Optionally, the above embodiment is a key analysis and classification step in this application, and its main function is to determine the interest similarity between different target accounts based on the attribute information of the common interest subsequence, and classify the accounts accordingly to finally obtain the classification results.

[0062] Optionally, interest similarity measures the degree to which two or more target accounts share interest content in their interest sequences. In this embodiment, interest similarity is calculated by analyzing the attribute information of all extracted common interest subsequences. Specifically, the attribute information generally includes the length of the longest common interest subsequence between two interest sequences and the number of all common interest subsequences. These attributes can reflect the similarity of interest depth (length of the longest common subsequence) and breadth (number of all common subsequences) between users.

[0063] Alternatively, various methods can be used to determine interest similarity. For example, a weighted average of the length of the longest common subsequence and the number of all common subsequences can be taken, or more complex statistical models can be used to integrate this information. Regardless of the method chosen, the core goal is to quantify the similarity in interest content between different target accounts for subsequent classification.

[0064] Optionally, based on the calculated interest similarity, accounts in the target account set can be categorized, with the goal of grouping accounts with similar interests into one category. This process typically involves setting a similarity threshold, above which accounts are considered to have similar interests and are grouped into the same category. Account categorization is the foundation of personalized recommendations and precision marketing in recommendation systems. By categorizing accounts with similar interests, banks can develop specific marketing strategies for each category of accounts. For example, for user groups with a common preference for financial products, specialized marketing campaigns or benefit distribution programs can be designed to increase customer engagement and satisfaction.

[0065] In the embodiments of the present application, in the above embodiments, through account classification, banks can identify user groups with specific interest patterns, thereby conducting more precise marketing activities and improving marketing efficiency and conversion rates. When banks can accurately predict and respond to user interests, they can provide more personalized and relevant services, thereby improving user experience and enhancing customer satisfaction and loyalty. By comparing new users or new products with classified user groups, they can quickly find users with similar interests, alleviate cold start problems, accelerate the integration of new users and the promotion of new products. The classification results help banks to reasonably allocate marketing resources, ensure that resources are invested in the most potential user groups, avoid waste of resources, and improve overall marketing effectiveness. Based on account classification and interest similarity, banks can make more informed decisions, such as product development, service optimization, customer relationship management, etc., to ensure that the decision-making process is based on sufficient user behavior data.

[0066] In steps S201 to S204 of the embodiment of the present application, by integrating multi-dimensional data to construct interest sequences, determine common interest subsequences, calculate comprehensive interest similarities and perform classification based on this, not only the accuracy of classification is improved, but also a more effective precision marketing tool is provided for banks, which can better meet user needs and improve user satisfaction, solve the technical problem of low classification accuracy of objects in related technologies, and achieve the technical effect of improving the classification accuracy of objects.

[0067] The above method of this embodiment is further introduced below.

[0068] As an optional implementation method, step S202 constructs an interest sequence of the target account based on historical transaction information, preference information and account line information, including: constructing interest points for the target account based on historical transaction information, preference information and account line information, wherein the interest points are used to represent the target account's evaluation behavior on the service item at the timestamp; sorting the interest points in order of the timestamps to obtain an interest sequence.

[0069] In this embodiment, the above embodiment is a key step in constructing the target account interest sequence, which forms a time series representation that can reflect the user's dynamic interests and behavior patterns by integrating historical transaction information, preference information and account line information.

[0070] Optionally, a point of interest is the basic unit that constitutes an interest sequence, representing the user's evaluation behavior on a certain service item at a specific point in time. When constructing points of interest, the following three types of information are mainly used: historical transaction information, including the user's transaction records in the banking system, such as the purchase of financial products, deposits, loans, remittances, etc., as well as transaction-related details such as transaction amount, time, product type, etc. Transaction behavior can intuitively reflect the user's interest and preference for a certain type of service or product. Preference information covers the user's behavioral data on other external platforms, such as social media activities, shopping platform browsing, news reading preferences, etc. The above information is crucial for building a comprehensive user portrait. Account bank information refers to the basic information and account status registered by the user in the bank.

[0071] Optionally, when constructing POIs, the above information can be integrated and processed, converting each user transaction or preference behavior into a four-tuple consisting of a user identifier (ID), an item ID, a rating (or evaluation value), and a timestamp. The rating can be the user's actual transaction amount, a product satisfaction rating, or a preference rating simulated based on preference information. The timestamp is used to record the specific time the POI occurred, facilitating subsequent time series analysis.

[0072] Optionally, once the points of interest have been constructed, the next step is to sort them by timestamp order to generate an interest sequence. This sorting process ensures the temporal continuity of the points of interest and is the basis for analyzing the evolving patterns of user interests. Through time series analysis, banks can not only identify users' persistent interest in specific services or products, but also capture changing trends in interest and potential preference shifts.

[0073] Optionally, all points of interest for each target account are sorted in ascending order by timestamp. The accuracy of timestamps is crucial to the validity of the interest sequence, ensuring that the time of occurrence of each point of interest is correct. The sorted points of interest are linked in chronological order to form a continuous sequence. Each point of interest is a snapshot of the dynamic changes in user interests, and the linked interest sequence is a complete history of user interest changes. The generated interest sequence is evaluated to ensure the integrity and logic of the sequence. For example, check for time jumps or repeated points of interest to ensure that the sequence accurately reflects the continuous changes in user interests.

[0074] In this embodiment of the present application, through the above steps, banks can generate interest sequences that reflect users' dynamic behaviors and changing interests, providing strong data support for subsequent user similarity calculations, identification of common interest subsequences, and formulation of precision marketing strategies. Constructing interest sequences is a key step in transforming users from static data points into dynamic behavior sequences, helping banks gain a deeper and more comprehensive understanding of user needs, thereby enhancing the level of personalized services and targeted marketing campaigns.

[0075] As an optional implementation method, step S203 determines a common interest subsequence from the interest sequence corresponding to the target account in the target account set, including: obtaining the interest subsequence in the interest sequence; matching the interest subsequences corresponding to different target accounts in the target account set to obtain rating difference data, wherein the rating difference data is used to indicate the degree of difference in ratings of different target accounts for the same service item; and determining the common interest subsequence based on the rating difference data.

[0076] In this embodiment, a common interest subsequence is determined from the interest sequence in the target account set. This process aims to quantify the consistency of interests of different users in the same or similar service items, thereby laying the foundation for subsequent similarity calculation and classification.

[0077] Optionally, a series of interest subsequences are obtained from the interest sequence of each target account. An interest subsequence is a fragment of an interest sequence, consisting of consecutive interest points that reflect the changes in the user's interest in a specific service item over a period of time. The purpose of obtaining interest subsequences is for subsequent matching and comparison. The acquisition of interest subsequences usually involves traversing the interest sequence of each account and extracting all possible interest subsequences. Considering the length and combination of the sequences, this process may involve a large amount of computing resources. Therefore, it may be necessary to design an efficient algorithm to reduce the computational complexity, such as using a sliding window technique to quickly generate interest subsequences.

[0078] Optionally, interest subsequences from different accounts in the target account set are matched to calculate rating difference data. Rating difference data is a metric that measures the consistency of different users' evaluations of the same service item, reflecting the similarity of user interests. During the matching process, the system checks whether two interest subsequences have the same points of interest (i.e., user, item, timestamp) and compares the ratings of these points of interest.

[0079] Optionally, for each pair of common POIs, the rating difference is calculated using a rating difference function (as shown in the formula). This rating difference function takes into account both the absolute and temporal differences in ratings. A rating difference threshold can be set to determine whether two ratings are considered "equal." If the rating difference is less than the rating difference threshold, it indicates a high degree of consistency in user evaluations of the item; otherwise, it indicates a high degree of rating discrepancy.

[0080] Optionally, based on the score difference data, a common interest subsequence is determined. Common interest subsequences are all interest subsequences whose score differences in the interest sequences of different accounts meet the set score difference threshold. This process usually involves the calculation of two aspects of attribute information: the length of the longest common interest subsequence (LCS for short). LCS is the longest subsequence among all common interest subsequences. Its length reflects the consistency of the depth of user interests and is one of the key indicators for measuring user similarity. The number of all common interest subsequences, this indicator reflects the consistency of the breadth of user interests, that is, how many service items the user has shown consistent interest in.

[0081] Alternatively, the process of determining the common interest subsequences is usually implemented through data structures such as dynamic programming, prefix tree (Trie) or suffix tree, which can effectively handle the comparison of a large number of interest subsequences, thereby quickly finding the LCS and all common interest subsequences.

[0082] In the embodiment of the present application, through the above steps, the bank can identify common patterns of user interests, which not only helps to deeply understand the common needs of user groups, but also provides data support for personalized services and precision marketing strategies. For example, by comparing the differences in ratings of different users for the same financial product, the bank can find out which products are most popular among user groups and which users are likely to be open to new products or services. In addition, identifying common interest subsequences can also help banks solve the "cold start" problem of new users or new products. By matching with the preference data of existing user groups, it can quickly recommend services that may be of interest to new users, thereby improving user experience and engagement.

[0083] As an optional implementation method, the interest sub-sequences corresponding to different target accounts in the target account set are matched to obtain matching results, including: matching the interest sub-sequences corresponding to the evaluation behaviors performed by different target accounts on the same service item, and determining the score difference data corresponding to the interest sub-sequences corresponding to the same service item of different target accounts.

[0084] In this embodiment, matching the interest subsequences corresponding to different target accounts in the target account set to obtain matching results and calculate the score difference data is a key step in refining the user interest similarity analysis.

[0085] Optionally, to obtain matching results, the system first needs to extract interest subsequences related to specific service items from each target account's interest sequence. This means that the system needs to identify all evaluation behaviors of each user for the same service item, which may include multiple transactions, evaluations, or interactions. For example, if a bank wants to analyze user interest in a certain type of investment product, it will find all points of interest related to the investment product from each user's interest sequence and then construct these points of interest into one or more interest subsequences.

[0086] Optionally, once the interest subsequences are extracted, these interest subsequences are then matched against each other to calculate rating difference data, which reflects the consistency of different users' evaluations of the same service item.

[0087] Optionally, for each service item, the system compares ratings from different users at the same or similar timestamps. Rating differences can be measured using absolute value difference, squared difference, or other statistical metrics to reflect the closeness of the ratings. For example, if two users rate the same investment product 4.5 and 4.7, respectively, and the rating difference threshold θ is set to 0.3, then the difference between the two ratings is less than the threshold, and the ratings are considered consistent.

[0088] Optionally, for each service item, the degree of difference in ratings across all users is tallied to generate rating difference data. This data aggregation process typically involves calculating statistics such as the mean, median, and standard deviation of rating differences for each item to comprehensively measure the consistency of user interest in the item.

[0089] Optionally, analyze rating variance data to determine which services have generated widespread interest among user groups and which users have highly similar interests and preferences. Services with small rating variances often indicate high user consistency, while services with large rating variances may indicate significant differences in preferences between users.

[0090] Optionally, based on rating difference data, we determine which interest subsequences can be considered common interest subsequences. This typically involves setting a rating difference threshold θ. If the difference in ratings between different users for the same service item is less than θ, the relevant interest points are considered to be of common interest and thus constitute a common interest subsequence. The length and number of common interest subsequences are important indicators of interest similarity between users, reflecting the commonality in their interest evolution patterns.

[0091] In the embodiment of the present application, through the above steps, the bank can not only identify users' general interest and preference consistency for specific service items, but also more accurately calculate the interest similarity between users based on the length and number of common interest subsequences. This allows the bank to identify service items with high rating consistency and optimize and adjust product or service combinations to better meet user needs. It can also develop more precise marketing strategies based on user groups with small rating differences, improving the effectiveness of marketing activities and customer satisfaction. It can also group users with similar rating behaviors to facilitate personalized customer service and subsequent optimization of the recommendation system. By analyzing rating difference data and common interest subsequences, the bank can improve its recommendation algorithm, enhancing the accuracy of recommendations and user experience.

[0092] In summary, matching interest subsequences and calculating rating difference data provides banks with a tool for in-depth analysis of user interest consistency and diversity, a key step in achieving precision marketing and optimizing services. Through this process, banks can more effectively manage and engage their target customer base based on data-driven insights.

[0093] Optionally, for a certain group, the length of the longest common interest subsequence and the number of all common subsequences are calculated to calculate the user similarity based on the interest sequence:

[0094] The recommendation task can be considered a utility function that indicates a user's potential interest in an item. Assume that User = {u1,u2,...,un} represents the set of users, Item = {t1,t2,...,tn} represents the set of items, and Ts = {s1,s2,...,sn} represents the set of all timestamps at which users evaluate items. To describe the interest sequence-based recommendation system, the following definitions are given: An interest point (IP) represents a user's evaluation of an item at a certain timestamp and consists of a user, item, rating, and timestamp. An interest sequence (IS) represents a sequence of a user's interest points arranged in chronological order by timestamp.

[0095] Based on the above two definitions, a user's rating history can be converted into an interest sequence. Unlike existing recommendation methods that use user ratings on public items, this method utilizes user interest sequences to analyze the evolution of their interests. Because interest sequences contain more semantic information than simple interest points, they can not only reflect the dynamic interests of users but also the evolution of their interests. To calculate user similarity based on interest sequences, we consider the length of the longest common subsequence and the total number of common subsequences that have been verified in classification problems.

[0096] As an optional embodiment, determining the common interest subsequence based on the score difference data includes: determining the interest subsequence corresponding to the score difference data being less than or equal to the score difference threshold as the common interest subsequence.

[0097] In this embodiment, determining common interest subsequences based on rating difference data is an important step in refining the user interest similarity analysis in the technical solution. The above process aims to identify subsequences in the user group that have similar evaluation behaviors for the same service item.

[0098] Optionally, before determining the common interest subsequences, rating difference data must first be calculated or obtained. Rating difference data is information that quantifies the degree of difference in ratings given to the same service item by different users. It consists of a series of rating difference values ​​that reflect the difference between user ratings and a certain standard or between each other.

[0099] Optionally, since different users may use different rating scales, all ratings need to be converted to the same scale before calculating the rating difference, usually by normalizing the ratings to the interval [0, 1]. Setting a rating difference threshold is used to determine whether the difference between two ratings is small enough to consider the users' evaluation behavior on the item to be similar.

[0100] Optionally, after obtaining the score difference data, the next step is to determine which interest subsequences can be classified as common interest subsequences. This process is based on the interest subsequences whose score difference data is less than or equal to a score difference threshold.

[0101] Optionally, all possible interest subsequences are extracted from each user interest sequence in the target account set. For the same service item, the subsequence consisting of the interest points evaluated by all users on the item is used as the comparison object. Each pair of interest subsequences is compared, and the score difference data therein is checked to determine whether this data is less than or equal to a threshold. If the conditions are met, it means that this interest subsequence is common to users u and v, that is, they show similar interest in this service item. If all the score difference data in the interest subsequence are less than or equal to θ, then this subsequence is marked as a public interest subsequence. This means that in this subsequence, all users' evaluation behavior on the service item is consistent, and there is no significant difference in the ratings. The above comparison is performed for all users, and all interest subsequences marked as public are collected. These subsequences represent common and consistent interest patterns in the user group and are an important basis for the subsequent calculation of user similarity and optimization of the recommendation system.

[0102] In the embodiment of the present application, through the above steps, the bank can quantify the similarity of interests between users, especially for those service items with high consistency, and can more accurately identify the target customer group. This has the following benefits for the bank's marketing strategy: the bank can formulate marketing activities that are more targeted at specific user groups based on public interest subsequences, such as pushing relevant new product information to users who often participate in specific types of financial products; the identification of public interest subsequences can help banks understand which products or services have gained widespread recognition among users, thereby optimizing product portfolios and improving customer satisfaction; by collecting and analyzing public interest subsequences, banks can more effectively segment users and provide personalized service and product recommendations to customers in different groups; for new users or new products, by comparing with the public interest subsequences of known user groups, their possible points of interest can be quickly determined, thereby alleviating the cold start problem in the recommendation system.

[0103] Optionally, to describe the longest common interest subsequence and all common interest subsequences in the recommendation problem, the following definitions are given:

[0104] Given a rating difference threshold θ and interest sequence subsequences obtained from two users’ interest sequences isu and isv respectively, if these two interest sequence subsequences constitute an interest sequence match of length j, they are considered matched if and only if they satisfy the following two conditions:

[0105]

[0106] In the above formula, The function that can be used to express the difference between the scores of user u and user v for the same item (service item) can be defined as the following formula:

[0107]

[0108] Among them, max(r u ) can be used to represent the maximum score of user u on the project, min(r u ) can be used to represent the minimum score of user u on the project, max(r v ) can be used to represent the maximum value of user v’s rating of an item, min(r v ) can be used to represent the minimum score of user v on the item.

[0109] Optionally, due to the different rating scales of different users (some users with loose requirements may give full marks to their favorite items, while users with strict requirements may only be willing to give 70% of the score to their favorite items), all user ratings should be normalized to the same scale [0,1]. If the difference in user ratings is less than the rating difference threshold θ, then the ratings of two users can be considered equal. A smaller rating difference threshold θ means a stricter similarity limit, but generally, overly strict similarity limits will limit the effectiveness of interest sequences in recommendation systems, because overly strict similarity limits will reduce the length of the longest common interest subsequence between users and also reduce the total number of common interest subsequences. Therefore, the rating difference threshold θ should be optimized based on the application's sensitivity to interest sequences.

[0110] As an optional embodiment, the attribute information includes target length information and quantity information of the longest common interest subsequence, wherein, before determining the interest similarity between different target accounts based on the attribute information of the common interest subsequence, and classifying the target accounts based on the interest similarity, the method also includes: obtaining the length information of the common interest subsequence; based on the length information, determining the longest common interest subsequence from the common interest subsequence, and obtaining the target length information of the longest common interest subsequence; obtaining the quantity information of the common interest subsequences between two target accounts in the target account set; the interest similarity includes a first interest similarity and a second interest similarity, wherein, based on the attribute information of the common interest subsequence, determining the interest similarity between different target accounts includes: determining the first interest similarity based on the target length information; determining the second interest similarity based on the quantity information; and determining the interest similarity based on the first interest similarity and the second interest similarity.

[0111] In this embodiment, the use of attribute information in the above embodiment is to more comprehensively evaluate the interest similarity between different target accounts, ensuring the effectiveness and accuracy of the recommendation algorithm. This process involves in-depth analysis of common interest subsequences, including extracting their length and number information, and calculating the first interest similarity and second interest similarity based on this information.

[0112] Optionally, the length of each subsequence is obtained from all calculated common interest subsequences. The length of an interest subsequence reflects the length of a user's continuous evaluation or interaction with a specific service item over time and is a quantitative indicator of the depth of interest. For example, two users may give similar reviews of the same investment product, indicating their sustained interest in the product and similar interest patterns.

[0113] Optionally, after obtaining the lengths of all common interest subsequences, the longest common interest subsequence can be identified and determined. LCS is the longest common subsequence in two or more interest sequences. It not only provides the most direct comparison of interest depth, but also reflects the consistency of interest evolution between users. The target length information, that is, the length of LCS, is the basic data for calculating the first interest similarity. In addition to the length of LCS, the number of all common interest subsequences must be counted, which is a quantitative indicator of interest breadth. The quantity information reflects the frequency of users' consistent evaluations on different service items. Even if these subsequences may be shorter, the large number also indicates the broad consistency of user interests.

[0114] Optionally, the calculation of the first interest similarity is based on the LCS target length information extracted above. Generally, the longer the length of the LCS, the higher the interest similarity between users. The first interest similarity can be calculated in a variety of ways, such as directly using the length of the LCS, or performing some form of normalization on it to make it more suitable as a similarity measure. The determination of the second interest similarity is based on the quantity information of the common interest subsequences. The increase in quantity information indicates that the consistency of interests between users is not limited to one or two items, but spans multiple service items, which helps to understand the similarity of user interests more comprehensively. The second interest similarity can be calculated by standardizing the number of common interest subsequences or using a probability model.

[0115] Optionally, after obtaining the first interest similarity and the second interest similarity, the two can be combined to more accurately determine the final interest similarity. This combination process may involve weight adjustment, that is, assigning different weights to the first interest similarity and the second interest similarity based on application needs and specific data conditions.

[0116] Optionally, with the help of the interest sequence matching defined above, all common interest subsequences and the longest common interest subsequence can be defined as follows:

[0117] Given two users’ interest sequences is u and is v , if there exists a set A for the sequence of interest is u and is v Any interest sequence matching ism between them always satisfies ism∈A, and for any element e∈A in set A, e always meets the definition of interest sequence matching in Definition 3, which is the interest sequence is u and is v If an interest sequence between them matches, then the set A is called the interest sequence is u and is v All common interest subsequences between . Among them, all common interest subsequences include empty interest sequence matches.

[0118] Given two users’ interest sequences is u and is v , and all common interest subsequences A between them, if there is an interest sequence matching ism∈A, |ism| represents the length of the interest sequence matching ism, so that for any interest sequence matching x∈A in A, the condition |ism|≥|x| is satisfied, then the interest sequence matching ism is the interest sequence is u and is v The longest common interest subsequence between .

[0119] Optionally, the longest common interest subsequence and the total common interest subsequence provide shared common information between user interest sequences. Intuitively, if two users have a longer longest common interest subsequence and a larger total common interest subsequence, then their interest sequences are more similar. The specific calculation of user similarity based on the longest common interest subsequence and the total common interest subsequence will be described later. The recommendation task addressed in this chapter can be considered a rating prediction problem based on user interest sequences. The utility of items to users is evaluated based on the similarity between user interest sequences.

[0120] Optionally, give two interest sequences extracted from the rating history of user u and user v is u and is v , where |is u |=m,|is v |=n, let ω be a (m+1)×(n+1) matrix, and the length of the longest common interest subsequence between user u and user v is expressed as |lc(u,v)|, which is expressed by the following formula:

[0121]

[0122] Among them, 0≤i≤m, 0≤j≤n, then |lc(u,v)|=ω[m,n].

[0123] Interest sequence is u and is v The total number of common interest subsequences is represented by |a(u,v)| and is expressed by the following formula:

[0124]

[0125] Among them, 0≤i≤m, 0≤j≤n, then |a(u,v)|=ω[m,n].

[0126] Alternatively, consider two interest sequences is u and is v, where the set of items in the interest sequence is represented as I = {C, D, E, F}. For the convenience of description, for any item it∈I, set the score r∈[0,5]. In addition, set the score difference threshold θ = 0.2. It can be seen that the interest sequence is u and is v The set of the longest common interest subsequence is {C->F, E->F}, so the length of the longest common interest subsequence can be obtained as |lcsis(u,v)|=2.

[0127] It should be noted that the calculation of the length of the longest common interest subsequence and the number of all common interest subsequences must take into account the differences in users' ratings of the same item. Get the position index x of the interest point in the interest sequence,

[0128] Make The specific description is as follows:

[0129]

[0130] Among them, 1≤x≤j.

[0131] To compare the similarity of two users' interest sequences, we first need to regularize the length of the longest common interest subsequence lc(u,v) and the total number of common interest subsequences a(u,v). Regularizing the length of the longest common interest subsequence lc(u,v) yields the user similarity based on the length of the longest common interest subsequence.

[0132]

[0133] The number of all common interest subsequences a(u,v) is regularized to obtain the user similarity based on the number of all common interest subsequences.

[0134]

[0135] Then, the parameter а is used to merge the user similarity based on the length of the longest common interest subsequence and the user similarity based on the number of all common interest subsequences to obtain a mixed user similarity based on interest sequences. Under normal circumstances, the similarity is very small and needs to be amplified using the parameter b.

[0136] sim is (u, v) = a × b × sim lc (u, v)+(1-a)×b×sim a (u, v)

[0137] As an optional embodiment, step S204 classifies the target accounts based on interest similarity, including: integrating interest similarity into the Pearson coefficient to obtain target interest similarity; and classifying the target accounts based on the target interest similarity.

[0138] In this embodiment, interest similarity is integrated into the Pearson coefficient, and target accounts are classified based on target interest similarity. This is a key step in achieving more accurate user grouping and personalized recommendations.

[0139] Alternatively, interest similarity can be used to quantify the consistency of interests between users by calculating the length of the longest common interest subsequence and the number of common interest subsequences. However, using these metrics alone may not fully reflect the relevance and preference consistency between users because they mainly focus on the direct similarity of user behavior and ignore the linear correlation or trend of ratings.

[0140] Alternatively, the Pearson Correlation Coefficient is a statistic used to measure the linear correlation between two variables. In recommendation systems, the Pearson Correlation Coefficient is often used to measure the linear correlation between user ratings of items. It takes into account factors such as the mean and variance of user ratings and can reflect the overall trends and patterns of user ratings.

[0141] Optionally, in order to combine the similarity of interest behavior and the correlation of scoring patterns, interest similarity can be integrated into the calculation of the Pearson coefficient to form a target interest similarity. Specifically, when calculating the target interest similarity, the following steps can be followed: Calculate the original Pearson correlation coefficient between users, which reflects the consistency of the scoring trend. Calculate the interest similarity based on the length of the longest common interest subsequence and the number of common interest subsequences. Use parameters to mix the original Pearson coefficient and interest similarity to form a new target interest similarity. This can be achieved through linear weighting, multiplication or other mathematical operations.

[0142] Optionally, based on the target interest similarity, a threshold can be set to determine which users can be classified into the same group. For example, if the target interest similarity exceeds a preset threshold, then these users can be regarded as a group with similar interests. Use clustering algorithms (such as K-means, hierarchical clustering, etc.) to classify the target accounts. These algorithms will automatically group similar accounts into the same group based on the similarity matrix between accounts. Target interest similarity, as a distance or similarity measure in the clustering process, can help the algorithm more accurately identify the consistency of interests between users.

[0143] Optionally, after the clustering process is complete, a series of user groups will be generated, each with users having a high degree of similarity in target interests. These groups can be further analyzed based on specific marketing objectives or product characteristics to determine more specific marketing strategies or product recommendations.

[0144] In the embodiment of this application, through the above steps, banks can more comprehensively assess the similarity between users based not only on the direct similarity of user interests and behaviors, but also on the consistency of the comprehensive scoring model. This enables banks to: improve the accuracy of user group segmentation, ensuring that users within each group have similar interests and preferences; provide more personalized service and product recommendations based on user groups, improving user experience and satisfaction; conduct targeted marketing for user groups with high interest similarity, reducing marketing costs and improving marketing efficiency and conversion rates; and by analyzing the common interests of different user groups, banks can better understand market demand and optimize product portfolios and service processes.

[0145] Optionally, the user similarity based on interest sequence is mixed with the similarity in the collaborative filtering recommendation system. This method can further improve the accuracy of measuring user similarity by using the features hidden in the interest sequence. For user-based recommendation systems, the Pearson correlation coefficient performs better than other user similarity measurement methods. The similarity based on interest sequence is integrated into the Pearson correlation coefficient:

[0146] sim(u,v)=pc(u,v)×F(sim IS (u,v))

[0147] pc(u,v) represents the Pearson correlation coefficient between user u and user v, and F(simIS(u,v)) represents the influence coefficient of the similarity based on interest sequence on user similarity:

[0148]

[0149] Among them, comm(u,v) represents the number of items that user u and user v have jointly evaluated, and total(u,v) represents the number of the union of all items evaluated by user u and user v.

[0150] As an optional embodiment, the method further includes: formulating corresponding marketing strategies for target accounts of different categories, and recommending service items to the target accounts according to the marketing strategies.

[0151] In this embodiment, corresponding marketing strategies may be formulated for the target accounts of different categories, and service items may be recommended to the target accounts according to the marketing strategies.

[0152] Optionally, after calculating the similarity between all users, sort the other users from highest to lowest based on their similarity to the target user. The top K users who have rated the project to be rated by the target user are selected as the target user's neighbors for a particular project. Based on the neighbor users' ratings of the target project and the similarity between the neighbor users and the target user, the target user's potential rating for the target project is predicted. The similarity between the selected customers who have not yet participated in the event and those who have participated can be assessed, and customized event benefits or targeted marketing can be quantitatively delivered to these target customers to achieve precision marketing goals.

[0153] Alternatively, developing and implementing corresponding marketing strategies for different target account categories is a key step in transforming interest-based user segmentation into real business value. This process allows banks to provide more personalized and targeted services based on the specific interests and behavior patterns of user groups.

[0154] Optionally, once the marketing strategy is developed, the bank can recommend the most relevant and attractive service items for each user category according to the strategy.

[0155] Optionally, after implementing a marketing strategy, banks should monitor user responses and engagement within each user category, collect feedback, and evaluate marketing effectiveness. This includes, but is not limited to, analyzing engagement with each service offering and identifying which offers perform best for specific user categories. User feedback can be collected through surveys, online reviews, or customer service interactions to understand the effectiveness of the marketing strategy. Based on the effectiveness evaluation results and customer feedback, the marketing strategy and recommended services can be adjusted to further improve customer satisfaction and engagement.

[0156] In the embodiment of this application, through the above steps, banks can transform interest-based user classification into actual marketing actions, which helps improve the accuracy and efficiency of marketing activities, reduce resource waste, and enhance customer stickiness and satisfaction. The advantage of this personalized marketing approach is that it can more directly respond to users' real needs and interests, thereby improving the market competitiveness of banking services and customer loyalty.

[0157] The technical solutions of the embodiments of the present application are illustrated below with reference to preferred implementation methods.

[0158] Currently, relevant recommendation systems are primarily based on collaborative filtering. While collaborative filtering recommendation methods offer the advantages of ease of use and strong versatility, they also suffer from several drawbacks and issues. A prerequisite for collaborative filtering recommendation methods to achieve effective recommendations is sufficient rating data within the system. However, due to data sparsity and the cold start problem, the effectiveness and scope of application of collaborative filtering recommendation methods are limited. Furthermore, in practical application systems, collaborative filtering technology faces scalability issues. As users interact with the system, new data is constantly generated. Similarities obtained by memory-based collaborative filtering techniques, or models and parameters learned by model-based collaborative filtering techniques, cannot be directly used and must be recalculated. Furthermore, although collaborative filtering recommendation methods have achieved significant development after years of research, with significant improvements in recommendation precision and accuracy, their accuracy remains suboptimal and cannot meet industry needs.

[0159] In addition, existing collaborative filtering ignores the user's behavior sequence, which is also related to improving the accuracy of recommendations. This is because each user's characteristics will produce a unique rating sequence, which can reflect and analyze the user's real dynamic interest evolution pattern.

[0160] To address the above issues, the present application proposes a method for applying interest-sequence-based collaborative filtering to bank marketing activities. This method extracts account features based on customer account characteristics and transaction characteristics, combined with customer information. This method incorporates a method based on interest-sequence user similarity, linearly blending it with the similarity in traditional collaborative filtering recommendation systems. This method is explored for bank marketing and equity issuance scenarios, and specialized scenario analysis is incorporated, providing an alternative solution for bank precision marketing. This method addresses the technical issue of low object classification accuracy in related technologies, achieving the technical effect of improving object classification accuracy.

[0161] Related recommendation methods based on publicly reviewed items cannot distinguish differences in user interest evolution patterns and migration probabilities. However, temporal relationships can reflect the evolution of user preferences and experiences, and to some extent, these evolution patterns can reveal implicit information that influences user experiences and interests. Furthermore, inspired by the analysis of spatiotemporal trajectory data in location-based recommendation systems, since user interests are not static but dynamic, the temporal order of interests in online recommendation systems may also have more semantic meaning than individual points of interest, and this semantic meaning can be used to analyze the evolution of users' dynamic interests. The dynamic evolution of interests can be analyzed using interest sequences that include the trajectory of user interest migration between items and the temporal sequence of user behavior. Instead, the correlation between users' interest evolution patterns is measured based on the length of the longest common interest subsequence and the number of all common interest subsequences between different users' interest sequences. Based on this correlation, the similarity of users' interest evolution patterns is derived. By linearly blending this similarity with the similarity used in traditional collaborative filtering recommendation systems, a new method for calculating user interest similarity based on the hybrid interest sequence similarity is designed. Based on the mixed similarity, the K users most similar to the target user are selected as neighbor users, and the possible rating of the target user on the project is predicted based on the ratings of the neighbor users on the project and the similarity between the target user and the neighbor users.

[0162] Figure 3 Flowchart of the application method of collaborative filtering based on interest sequence in bank marketing activities according to the embodiment of the present application. Figure 3 As shown, the method may include the following steps:

[0163] Step S301: Acquire the customer's historical transaction information, customer social preference information, and customer bank information.

[0164] In this embodiment, it is a data collection module, which specifically includes three submodules: historical transaction information collection, customer social preference information collection, and internal bank information collection module.

[0165] Optionally, the bank's information collection is based on the bank's customer product situation and transaction situation. Due to sparse social relationships, it is necessary to first use expert rules to extract customers with high stickiness within the bank, such as remittance customers, customers who have purchased investment products multiple times, and users who have a good remittance relationship with the above customers.

[0166] Alternatively, the collection of customer social preference information primarily aims to capture browsing and transaction data on commonly used consumer apps, including those for finance, retail, and dining. Given the increasing awareness of data assets among companies, the collection of external social relationships should be considered primarily through the lens of federated learning. Initial customer clustering is performed on the bank using input such as customer ID, user spending history, and current investment status. The bank then combines user postings, spending history, browsing history, and following information from various social and consumer apps to identify preference clusters and simulate user ratings.

[0167] Step S302: determining a set of user interest points based on historical transaction information, customer social preference information, and customer internal bank information.

[0168] In this embodiment, the length of the longest common interest subsequence and the number of all common subsequences are calculated for a group to calculate the user similarity based on the interest sequence. The recommendation task can be considered as a utility function that indicates a user's potential interest in a certain item. Assume that User = {u1, u2, ..., u n} represents the user's collection, Item={t1,t2,...,t n} represents the set of items, Ts={s1,s2,...,s n} represents the set of all timestamps of a user's item reviews. To describe the recommendation system based on interest sequences, we first give the following definitions: An interest point (IP) represents a user's evaluation of an item at a certain timestamp and consists of the user, item, rating, and timestamp. An interest sequence (IS) represents a sequence of a user's interest points arranged in chronological order according to their timestamps.

[0169] Based on the above two definitions, the user's rating history can be converted into an interest sequence. Different from the existing recommendation method that uses users' ratings on public items, the user's interest sequence is used here to analyze the user's interest evolution characteristics. Because interest sequences have more semantic information than simple interest points, interest sequences can not only express users' dynamic interests, but also reflect the evolution characteristics of users' interests. In order to calculate the user similarity based on interest sequences, the length of the longest common subsequence and the number of all common subsequences that have been verified in the classification problem are considered. In order to describe the longest common interest subsequence and all common interest subsequences in the recommendation problem, the following definitions are given:

[0170] Interest sequence matching: Given a score difference threshold θ and interest sequence subsequences obtained from two users’ interest sequences isu and isv respectively, if these two interest sequence subsequences constitute an interest sequence match of length j, they are considered to match if and only if they satisfy the following two conditions:

[0171]

[0172] In the above definition, is a function that calculates the difference between the ratings of user u and user v on the same item. It can be defined as the following formula:

[0173]

[0174] Because different users have different rating scales (some users with looser requirements may give their favorite items a full score, while more demanding users may only be willing to give their favorite items a score of 70%), all user ratings should be normalized to the same scale [0, 1]. If the difference in user ratings is less than the rating difference threshold θ, then the two users' ratings are considered equal. A smaller rating difference threshold θ means a stricter similarity limit, but generally, overly strict similarity limits will limit the effectiveness of interest sequences in recommendation systems. This is because overly strict similarity limits will reduce the length of the longest common interest subsequence between users and also reduce the total number of common interest subsequences. Therefore, the rating difference threshold should be optimized based on the application's sensitivity to interest sequences.

[0175] Step S303 : Calculate the length of the longest common interest subsequence and the number of all common interest subsequences.

[0176] In this embodiment, with the help of the interest sequence matching defined above, all common interest subsequences and the longest common interest subsequence can be defined as follows:

[0177] All common interest subsequences. Given two user interest sequences isu and isv, if there exists a set A such that for any interest sequence match ism between the interest sequences isu and isv, ism∈A is always satisfied, and for any element e∈A in set A, e always meets the definition of interest sequence match in the above definition and is an interest sequence match between the interest sequences isu and isv, then set A is called all common interest subsequences between the interest sequences isu and isv. All common interest subsequences include empty interest sequence matches.

[0178] The longest common interest subsequence, given two users' interest sequences isu and isv, and all the common interest subsequences A between them, if there exists an interest sequence matching ism∈A, |ism| represents the length of the interest sequence matching ism, so that for any interest sequence matching x∈A in A, the condition |ism|≥|x| is satisfied, then the interest sequence matching ism is the longest common interest subsequence between the interest sequences isu and isv.

[0179] The longest common interest subsequence and the total common interest subsequence provide shared common information between user interest sequences. Intuitively, if two users have a longer longest common interest subsequence and a larger total common interest subsequence, then their interest sequences are more similar. The specific calculation of user similarity based on the longest common interest subsequence and the total common interest subsequence will be described later. The recommendation task addressed in this chapter can be considered a rating prediction problem based on user interest sequences. The utility of items to users is evaluated based on the similarity between user interest sequences.

[0180] Given two interest sequences isu and isv extracted from the rating histories of user u and user v, where |isu| = m and |isv| = n, let ω be a (m+1)×(n+1) matrix. The length of the longest common interest subsequence of user u and user v is represented by |lc(u,v)|, which is expressed by the following formula:

[0181]

[0182] Among them, 0≤i≤m, 0≤j≤n, then |lc(u,v)|=ω[m,n].

[0183] Interest sequence is u and is v The total number of common interest subsequences is represented by |a(u,v)| and is expressed by the following formula:

[0184]

[0185] Among them, 0≤i≤m, 0≤j≤n, then |a(u,v)|=ω[m,n].

[0186] Table 1 Two sequences of interest

[0187] T1 T2 T3 T4 <![CDATA[is u ]]> (C,2.5) (E,3.0) (D,4.5) (F,0.5) <![CDATA[is v ]]> (E,4.0) (D,2.5) (C,3.5) (F,1.5)

[0188] Optionally, Table 1 is a table of two interest sequences according to an embodiment of the present application. As shown in Table 1, consider the two interest sequences isu and isv in the table, where the set of items in the interest sequence is represented as I = {C, D, E, F}. For the sake of convenience of description, for any item it∈I, the score r∈[0,5] is set. In addition, the score difference threshold θ is set to 0.2. It can be seen that the set of the longest common interest subsequences of the interest sequences isu and isv is {C->F, E->F}, so the length of the longest common interest subsequence can be obtained as |lcsis(u,v)| = 2.

[0189] Step S304: determine the similarity of the interest sequence.

[0190] In this embodiment, the calculation of the length of the longest common interest subsequence and the number of all common interest subsequences must take into account the differences in users' ratings of the same item. Get the position index x of the interest point in the interest sequence,

[0191] Make The specific description is as follows:

[0192]

[0193] Among them, 1≤x≤j.

[0194] To compare the similarity of two users' interest sequences, we first need to regularize the length of the longest common interest subsequence lc(u,v) and the total number of common interest subsequences a(u,v). Regularizing the length of the longest common interest subsequence lc(u,v) yields the user similarity based on the length of the longest common interest subsequence.

[0195]

[0196] The number of all common interest subsequences a(u,v) is regularized to obtain the user similarity based on the number of all common interest subsequences.

[0197]

[0198] Then, the parameter а is used to merge the user similarity based on the length of the longest common interest subsequence and the user similarity based on the number of all common interest subsequences to obtain a mixed similarity of users based on interest sequences. Under normal circumstances, the similarity is very small and needs to be amplified using the parameter b.

[0199] sim is (u, v) = a × b × sim lc (u, v)+(1-a)×b×sim a (u, v)

[0200] Step S305: perform similarity fitting.

[0201] In this embodiment, user similarity based on interest sequences is combined with similarity in collaborative filtering recommendation systems. This method can further improve the accuracy of measuring user similarity by utilizing features hidden in interest sequences. For user-based recommendation systems, the Pearson correlation coefficient performs better than other user similarity metrics. By integrating similarity based on interest sequences into the Pearson correlation coefficient:

[0202] sim(u,v)=pc(u,v)×F(sim IS(u,v))

[0203] pc(u,v) represents the Pearson correlation coefficient between user u and user v, and F(simIS(u,v)) represents the influence coefficient of the similarity based on interest sequence on user similarity:

[0204]

[0205] Among them, comm(u,v) represents the number of items that user u and user v have jointly evaluated, and total(u,v) represents the number of the union of all items evaluated by user u and user v.

[0206] Step S306: Conduct potential customer marketing.

[0207] In this embodiment, after calculating the similarity between all users, other users are sorted from highest to lowest based on their similarity to the target user. The top K users who have rated the project to be rated by the target user are selected as the target user's neighbors for a particular project. Based on the neighbor users' ratings of the target project and the similarity between the neighbor users and the target user, the target user's potential rating for the target project is predicted. The similarity between the selected customers who have not yet participated in the event and those who have participated is evaluated, and customized event benefits or targeted marketing can be quantitatively sent to these target customers to achieve the goal of precision marketing.

[0208] In this embodiment, leveraging the generally long-term retention of bank users, this method can identify more points of interest. By calculating the similarity of interest sequences, the accuracy of the recommendation algorithm can be effectively improved. By setting a similarity threshold, specific marketing target groups can be identified, allowing for precise targeting and improved customer retention.

[0209] Alternatively, recommendation methods that consider time can also mix the impact of time sequence on the recommendation system, and many methods to improve the recommendation system have been proposed. However, they do not consider the impact of the entire time series on the recommendation system, and the effect is relatively poor compared to calculations based on interest sequences.

[0210] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0211] Example 2

[0212] The present application also provides an object classification device. It should be noted that the object classification device of the present application can be used to execute the object classification method provided by the present application. The object classification device provided by the present application is introduced below. According to the present application, a device for implementing the above-mentioned object classification method is also provided. Figure 4 is a schematic diagram of an object classification device provided according to an embodiment of the present application, such as Figure 4 As shown, the object classification device 400 includes: an acquisition unit 402, a construction unit 404, a determination unit 406, and a classification unit 408. The acquisition unit 402 is used to acquire historical transaction information, preference information, and account line information corresponding to a target account in a target account set; the construction unit 404 is used to construct an interest sequence for the target account based on the historical transaction information, preference information, and account line information; the determination unit 406 is used to determine a common interest subsequence from the interest sequences corresponding to the target account in the target account set; and the classification unit 408 is used to determine a common interest subsequence from the interest sequences corresponding to the target account in the target account set.

[0213] It should be noted that the acquisition unit 402, construction unit 404, determination unit 406, and classification unit 408 correspond to steps S201 to S204 in Example 1. The four modules and corresponding steps implement the same examples and application scenarios, but are not limited to the contents disclosed in the above-mentioned Example 1. It should be noted that the above-mentioned modules or units can be hardware components or software components stored in a memory (e.g., memory 104) and processed by one or more processors (e.g., processors 102a, 102b, ..., 102n). The above-mentioned modules can also be part of the device and can be run in the computer terminal 10 provided in Example 1.

[0214] Example 3

[0215] An embodiment of the present application may provide an electronic device, Figure 5 This is a structural block diagram of an electronic device according to an embodiment of the present application. Figure 5 As shown, the electronic device may include: one or more ( Figure 5 Only one is shown) processor 502, memory 504, storage controller, and peripheral interface, wherein the peripheral interface is connected to the radio frequency module, audio module and display.

[0216] Among them, the memory can be used to store software programs and modules, such as program instructions / modules corresponding to the methods and devices in the embodiments of the present application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, that is, implementing the above-mentioned method. The memory may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include a memory remotely arranged relative to the processor, and these remote memories may be connected to the terminal via a network. Examples of the above-mentioned network include but are not limited to the Internet, an intranet, a local area network, a mobile communication network and a combination thereof.

[0217] The processor may call the information and application programs stored in the memory through the transmission device to execute the above steps.

[0218] It can be understood by those skilled in the art that Figure 5 The structure shown is for illustration only, and the electronic device may also be a terminal device such as a smart phone (such as an Android phone, an iOS phone, etc.), a tablet computer, a PDA, a mobile Internet device (Mobile Internet Devices, abbreviated as MID), a tablet computer (Pad Device, abbreviated as PAD), etc. Figure 5 It does not limit the structure of the above electronic device. For example, the electronic device may also include Figure 5 More or fewer components (such as network interfaces, display devices, etc.) shown in, or with Figure 5 Different configurations shown.

[0219] A person skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing the hardware related to the terminal device through a program, and the program can be stored in a computer-readable storage medium, which may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0220] Example 4

[0221] The embodiment of the present application further provides a storage medium. Optionally, in this embodiment, the storage medium can be used to store the program code executed by the object classification method provided in the first embodiment.

[0222] Optionally, in this embodiment, the storage medium may be located in any computer terminal in a computer terminal group in a computer network, or in any mobile terminal in a mobile terminal group.

[0223] The present application also provides a computer program product, which, when executed on a data processing device, is adapted to execute the steps of the method for classifying an object.

[0224] The serial numbers of the above embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0225] In the above embodiments of the present application, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0226] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0227] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0228] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0229] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, ROM, RAM, mobile hard drives, magnetic disks or optical disks.

[0230] The above is only a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for classifying an object, characterized in that: include: Obtaining historical transaction information, preference information, and account line information corresponding to a target account in a target account set, wherein the historical transaction information is used to indicate the historical transaction behavior of the target account in the service system, and the account line information is used to indicate the attributes of the target account registered in the service system; constructing an interest sequence of the target account based on the historical transaction information, the preference information, and the account line information, wherein the interest sequence is used to represent the changing state of the target account's interest content over time; Determining a common interest subsequence from the interest sequences corresponding to the target accounts in the target account set, wherein the common interest subsequence is the same interest subsequence in the interest sequences of different target accounts; Based on the attribute information of the common interest subsequence, the interest similarity between different target accounts is determined, and based on the interest similarity, the target accounts are classified to obtain a classification result, wherein the interest similarity is used to indicate the degree of similarity between different target accounts to the same interest content.

2. The method according to claim 1, characterized in that Constructing an interest sequence of the target account based on the historical transaction information, the preference information, and the account bank information, including: Constructing points of interest for the target account based on the historical transaction information, the preference information, and the account line information, wherein the points of interest are used to represent the evaluation behavior of the target account on the service item at a timestamp; The interest points are sorted according to the order of the timestamps to obtain the interest sequence.

3. The method according to claim 2, characterized in that Determining a common interest subsequence from the interest sequences corresponding to the target accounts in the target account set includes: Obtaining the interest subsequence in the interest sequence; Matching the interest subsequences corresponding to different target accounts in the target account set to obtain rating difference data, wherein the rating difference data is used to indicate the degree of difference in ratings given by different target accounts to the same service item; The common interest subsequence is determined based on the score difference data.

4. The method according to claim 3, characterized in that Matching the interest subsequences corresponding to different target accounts in the target account set to obtain matching results includes: The interest sub-sequences corresponding to the evaluation behaviors performed by different target accounts on the same service item are matched, and the rating difference data corresponding to the interest sub-sequences corresponding to the same service item by different target accounts are determined.

5. The method according to claim 4, characterized in that Determining the common interest subsequence based on the score difference data includes: The interest subsequence corresponding to the score difference data being less than or equal to the score difference threshold is determined as the common interest subsequence.

6. The method according to claim 1, characterized in that The attribute information includes target length information and quantity information of the longest common interest subsequence. Before determining interest similarities between different target accounts based on the attribute information of the common interest subsequence and classifying the target accounts based on the interest similarities, the method further includes: Obtaining the length information of the common interest subsequence; determining the longest common interest subsequence from the common interest subsequences based on the length information, and obtaining the target length information of the longest common interest subsequence; Obtaining the quantity information of the common interest subsequences between two target accounts in the target account set; The interest similarity includes a first interest similarity and a second interest similarity, wherein determining the interest similarity between different target accounts based on the attribute information of the common interest subsequence includes: determining the first interest similarity based on the target length information; determining the second interest similarity based on the quantity information; The interest similarity is determined based on the first interest similarity and the second interest similarity.

7. The method according to claim 6, characterized in that Classifying the target accounts based on the interest similarity includes: Integrating the interest similarity into the Pearson coefficient to obtain the target interest similarity; The target accounts are classified based on the target interest similarity.

8. The method according to claim 1, characterized in that The method further comprises: Corresponding marketing strategies are formulated for the target accounts of different categories, and service items are recommended for the target accounts according to the marketing strategies.

9. An object classification device, characterized in that: include: an acquisition unit, configured to acquire historical transaction information, preference information, and account line information corresponding to a target account in a target account set, wherein the historical transaction information is used to represent historical transaction behavior of the target account in the service system, and the account line information is used to represent attributes of the target account registered in the service system; a constructing unit, configured to construct an interest sequence of the target account based on the historical transaction information, the preference information, and the account line information, wherein the interest sequence is used to represent a change state of the target account's interest content over a time dimension; a determining unit, configured to determine a common interest subsequence from the interest sequences corresponding to the target accounts in the target account set, wherein the common interest subsequence is the same interest subsequence in the interest sequences of different target accounts; A classification unit is used to determine the interest similarity between different target accounts based on the attribute information of the common interest subsequence, and classify the target accounts based on the interest similarity to obtain a classification result, wherein the interest similarity is used to indicate the degree of similarity between different target accounts to the same interest content.

10. A processor, characterized in that: The processor is configured to run a program, wherein the program, when run by the processor, executes the object classification method according to any one of claims 1 to 8.

Images