Personalized data recommendation method and device for leasing business and electronic equipment

Abstract

The invention provides a rental service personalized data recommendation method and device and electronic equipment, and belongs to the technical field of rental services. The method comprises the following steps: extracting basic information and behavior data of users from a platform database, subdividing user groups by using a clustering algorithm based on the basic information and behavior data of the users, and establishing a detailed user portrait for each user; selecting a target algorithm model according to the user portrait and the features of the leased product, training, and optimizing model parameters through cross validation; and when the user accesses, the recommendation result is dynamically adjusted according to the behavior data of the current session, and multi-channel pushing is realized. Through fusion of user portraits, behavior analysis and an intelligent recommendation algorithm, accurate matching and dynamic optimization of personalized recommendation in the leasing business are realized, and the user satisfaction and the platform conversion efficiency are significantly improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of leasing business, and particularly relates to a leasing business personalized data recommendation method and device and electronic equipment. BACKGROUND

[0002] With the rapid development of the sharing economy and digital services, leasing business has been widely used in housing, vehicles, office equipment, electronic products and other fields.

[0003] Traditional leasing platforms often use general recommendation strategies or simple hotness sorting methods when recommending products to users, which cannot meet the personalized needs of users, resulting in low matching efficiency, poor user experience, low conversion rate and other problems. Therefore, how to achieve accurate recommendation through big data and intelligent algorithm technology has become the key to improving the competitiveness of leasing platforms. SUMMARY

[0004] In order to solve at least one of the technical problems in the background art, the application provides a leasing business personalized data recommendation method, which realizes accurate matching and dynamic optimization of personalized recommendation in leasing business by fusing user portrait, behavior analysis and intelligent recommendation algorithm, and significantly improves user satisfaction and platform conversion efficiency.

[0005] The technical scheme adopted by the application is as follows:

[0006] The first aspect embodiment of the application provides a leasing business personalized data recommendation method, comprising:

[0007] Extracting basic information and behavior data of users from a platform database, using a clustering algorithm to subdivide user groups based on the basic information and behavior data of users, and establishing a detailed user portrait for each user;

[0008] Selecting a target algorithm model according to the user portrait and the characteristics of leasing products and training the target algorithm model, and optimizing model parameters through cross-validation;

[0009] Adjusting the recommendation result dynamically according to the behavior data of the current session when the user accesses, and realizing multi-channel push.

[0010] According to the method for recommending personalized data of a leasing service provided by the first aspect of the application, by systematically integrating basic information of a user, historical behavior and real-time interaction data, a highly representative user portrait system is constructed, and a recommendation model is dynamically selected and optimized based on the portrait result, so that accurate and intelligent leasing product recommendation services are realized. First, in the data collection and user grouping stage, the method extracts basic attribute information and multi-dimensional behavior data of the user from the platform database, and uses K-means or hierarchical clustering algorithm to subdivide the user groups, which not only improves the scientificity of user classification, but also lays a solid foundation for subsequent personalized recommendation. Secondly, in the model training and parameter optimization stage, the method combines the user portrait label and the key features of the leasing product to construct the user-item interaction matrix, and introduces collaborative filtering, content-based recommendation or hybrid model for training, and continuously adjusts the model parameters through cross-validation mechanism, which significantly improves the accuracy and generalization ability of the recommendation system. Finally, in the recommendation execution and pushing stage, the method can update the recommendation result in real time according to the current session behavior of the user, such as search keywords, browsing path, click preference, etc., and reach the user through various channels such as home page recommendation position, APP push, email notification, etc., which enhances the immediacy and diversity of the recommendation. Overall, the method effectively solves the problem of single recommendation strategy and low matching degree of traditional leasing platform, significantly improves the user experience, recommendation conversion rate and platform operation efficiency.

[0011] According to an embodiment of the application, the basic information and behavior data of the user are extracted from the platform database, the user groups are subdivided based on the basic information and behavior data of the user using clustering algorithm, and a detailed user portrait is established for each user, specifically:

[0012] Collect basic attribute information and historical behavior data of the user, including age, gender, occupation, browsing record, click rate, collection and order record;

[0013] The user is grouped by using K-means or hierarchical clustering algorithm, and user subgroups with similar behavior patterns and preferences are identified;

[0014] Combining the group characteristics and individual behavior trajectory of the user, a user portrait containing multi-dimensional labels of interest preference, consumption ability and geographical location tendency is constructed.

[0015] According to an embodiment of the application, the target algorithm model is selected and trained according to the user portrait and the characteristics of the leasing product, and the model parameters are optimized through cross-validation, specifically:

[0016] Based on the multi-dimensional labels in the user portrait and the attribute information of the leasing product, a user-item interaction feature matrix is constructed;

[0017] By combining collaborative filtering, content-based recommendation, or hybrid recommendation algorithms, a target recommendation model suitable for the rental business scenario is selected.

[0018] Cross-validation was used to fine-tune the model parameters, and the model performance was measured by accuracy, recall, and AUC to ensure the accuracy of the recommendation results.

[0019] According to one embodiment of this application, the step of dynamically adjusting the recommendation results based on the behavioral data of the current session when a user accesses the site, and implementing multi-channel push notifications, specifically includes:

[0020] Real-time collection of user behavior data in the current session, including search keywords, pages viewed, dwell time, and click preferences;

[0021] The recommendation model is updated in a lightweight manner based on an online learning mechanism, and the recommendation list is dynamically adjusted to match the user's real-time interests.

[0022] The updated recommendation results will be pushed through multiple channels, including homepage recommendations, emails, mobile app notifications, and SMS.

[0023] According to one embodiment of this application, the method further includes:

[0024] Introduce a user feedback mechanism, allowing users to rate, like, favorite, or mark as uninterested in the recommendation results;

[0025] Negative and positive samples are constructed based on user feedback data to optimize the training and iteration of the recommendation model.

[0026] According to one embodiment of this application, the method further includes:

[0027] Regularly synchronize and update user behavior data and rental product information from the platform database to ensure the timeliness and completeness of the data on which the recommendation system relies;

[0028] Historical behavior data is weighted based on a time decay function.

[0029] According to one embodiment of this application, the method further includes:

[0030] Analyze users' historical behavior and real-time interaction data to identify potential abnormal operations or malicious order-brushing behavior;

[0031] Using techniques such as graph neural networks or isolated forest algorithms, user behavior graphs are constructed and suspicious nodes are identified.

[0032] A second aspect of this application provides a personalized data recommendation device for rental services, comprising:

[0033] The classification module is suitable for extracting basic information and behavior data of the user from a platform database, subdividing a user group based on the basic information and behavior data of the user using a clustering algorithm, and establishing a detailed user portrait for each user;

[0034] The parameter optimization module is suitable for selecting a target algorithm model and training the target algorithm model according to the user portrait and characteristics of a rental product, and optimizing model parameters through cross validation.

[0035] The pushing module is suitable for dynamically adjusting a recommendation result according to behavior data of a current session when the user accesses, and implementing multi-channel pushing.

[0036] The third aspect of the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor executes the program to realize the method for recommending personalized data of a rental service in any embodiment of the first aspect.

[0037] The present application also provides a non-volatile computer storage medium having computer executable instructions stored thereon, and the computer program is executed by the processor to realize the method for recommending personalized data of a rental service in any embodiment of the first aspect. BRIEF DESCRIPTION OF DRAWINGS

[0038] The accompanying drawings, which are included to provide a further understanding of the present application, constitute a part of the present application, and the illustrative embodiments of the present application and their description serve to explain the present application, and do not constitute improper limitations on the present application. In the drawings:

[0039] Figure 1 The flowchart of the method for recommending personalized data of a rental service provided by the embodiments of the present application is shown;

[0040] Figure 2 The structural diagram of the device for recommending personalized data of a rental service provided by the embodiments of the present application is shown;

[0041] Figure 3 The structural diagram of the electronic device provided by the embodiments of the present application is shown.

[0042] Reference signs:

[0043] 110, classification module; 120, parameter optimization module; 130, pushing module;

[0044] 810, processor; 820, communication interface; 830, memory; 840, communication bus. DETAILED DESCRIPTION

[0045] In order to more clearly explain the overall concept of the present application, the following will be described in detail in an exemplary manner with reference to the accompanying drawings.

[0046] In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application. However, the present application can be practiced without the specific details. In other instances, well-known methods have not been described in detail in order not to unnecessarily obscure aspects of the present application. It is to be understood that this application can assume other different embodiments and its aspects and features can be combined except where otherwise explicitly stated.

[0047] In this application, unless otherwise explicitly specified and limited, the first feature is "on" or "under" the second feature, which can be direct contact between the first and second features, or indirect contact between the first and second features through an intermediate medium. In the description of the specification, the description of the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples.

[0048] As shown in Figure 1 The first aspect of the present application provides a method for recommending personalized data of rental business, comprising:

[0049] Step 100, extracting the basic information and behavior data of the user from the platform database, using clustering algorithm to subdivide the user group based on the basic information and behavior data of the user, and establishing detailed user portrait for each user.

[0050] Step 200, selecting target algorithm model according to the user portrait and rental product characteristics and training, and optimizing model parameters through cross-validation.

[0051] Step 300, dynamically adjusting the recommendation result according to the behavior data of the current session when the user visits, and realizing multi-channel push.

[0052] In step 100, the "basic information" in this step usually includes static attributes such as age, gender, occupation, registration information of the user; "behavior data" covers the operation track of the user on the rental platform, such as browsing record, click rate, collection, order, rating and other dynamic interactive information. "Clustering algorithm" refers to the method of automatically classifying data with similar characteristics, such as K-means or hierarchical clustering; "user portrait" is a structured label system constructed based on the above information, which is used to describe the interest preference, consumption ability, geographical location tendency and other multi-dimensional characteristics of the user.

[0053] By integrating the user's basic attributes and historical behavior data, and using clustering algorithms to finely group users, different types of user groups and their common needs can be effectively identified, thus avoiding the "one-size-fits-all" problem in the recommendation process. The user portrait system built on this basis not only improves the depth of user understanding, but also provides accurate data basis for subsequent personalized recommendation models, enhancing the explainability and adaptability of the recommendation system.

[0054] In step 200, "user portrait" is one of the input features, together with "rental product features" (such as rent, lease period, category, degree of new and old, etc.) to form the input variables of the recommendation model. The "target algorithm model" can be collaborative filtering, content-based recommendation algorithm, or a hybrid recommendation model combining multiple strategies. "Cross-validation" is a technique for evaluating and optimizing model performance, which divides the training set into multiple subsets for repeated training and validation to improve the generalization ability and stability of the model.

[0055] This step combines user portraits with rental product attributes to build a user-item interaction feature matrix, enabling the recommendation model to have stronger personalized modeling capabilities and significantly improving the relevance and matching degree of the recommendation results. Meanwhile, using cross-validation to optimize model parameters improves the robustness of the model and reduces the risk of overfitting, ensuring that the recommendation system has good stability and promotion ability in actual deployment, and is suitable for intelligent recommendation needs in large-scale rental business scenarios.

[0056] In step 300, "current session behavior data" refers to real-time operation logs generated by users during this visit, such as search keywords, page dwell time, click path, temporary collection, etc.; "dynamic adjustment of recommendation results" refers to real-time updating of the recommendation model output based on online learning mechanisms to reflect the immediate interest changes of users; "multi-channel push" refers to delivering recommended content to users through various ways such as homepage recommendation positions, APP notifications, SMS, email, etc.

[0057] This step introduces real-time behavior capture and online learning mechanisms, enabling the recommendation system to quickly respond to users' immediate interest changes and improve the timeliness and personalization level of recommendations. At the same time, through multi-channel push strategies, users can receive rental product information that matches their current needs on both web and mobile platforms, further enhancing user experience and platform stickiness, and improving conversion rates and user satisfaction.

[0058] According to the method for recommending personalized data of a leasing service provided in the first aspect of the present application, a highly representative user portrait system is built by systematically integrating user basic information, historical behavior and real-time interaction data, and a recommendation model is dynamically selected and optimized based on the portrait result, so as to realize accurate and intelligent leasing product recommendation service. First, in the data collection and user grouping stage, the method extracts the basic attribute information and multi-dimensional behavior data of the user from the platform database, and uses K-means or hierarchical clustering algorithm to subdivide the user groups, which not only improves the scientificity of user classification, but also lays a solid foundation for subsequent personalized recommendation. Secondly, in the model training and parameter optimization stage, the method combines the user portrait label and the key features of the leasing product to build a user-item interaction matrix, and introduces collaborative filtering, content-based recommendation or hybrid model for training, and continuously adjusts the model parameters through cross-validation mechanism, which significantly improves the accuracy and generalization ability of the recommendation system. Finally, in the recommendation execution and pushing stage, the method can update the recommendation result in real time according to the current session behavior of the user, such as search keywords, browsing path, click preference, etc., and reach the user through various channels such as home page recommendation position, APP push, email notification, etc., which enhances the immediacy and diversity of the recommendation. Overall, the method effectively solves the problem of single recommendation strategy and low matching degree of traditional leasing platform, significantly improves the user experience, recommendation conversion rate and platform operation efficiency.

[0059] In some embodiments of the present application, the basic information and behavior data of the user are extracted from the platform database, the user groups are subdivided based on the basic information and behavior data of the user using clustering algorithm, and a detailed user portrait is established for each user, specifically:

[0060] The basic attribute information and historical behavior data of the user are collected, including age, gender, occupation, browsing record, click rate, collection and order record;

[0061] The K-means or hierarchical clustering algorithm is used to group the users, and the user subgroups with similar behavior patterns and preferences are identified;

[0062] The user portrait containing multi-dimensional labels of interest preference, consumption ability and geographical location tendency is constructed based on the group characteristics and individual behavior trajectory of the user.

[0063] In this embodiment, first, the basic information and behavior data of the user are extracted from the platform database. These basic information includes the static attributes of the user such as age, gender, occupation, and the behavior data covers the dynamic interaction information of the user on the platform such as browsing record, click rate, collection list, historical order, etc. These data constitute the basis of user portrait construction, reflecting the long-term preference and short-term interest of the user.

[0064] By collecting the above multi-dimensional user data, the system can more comprehensively understand the user's behavior patterns and potential needs. Compared with relying on a single dimension (such as using only transaction data), the fusion of basic attributes and behavior trajectories significantly improves the accuracy of user preference description, providing reliable data support for subsequent precise recommendation.

[0065] Next, an unsupervised learning algorithm such as K-means or hierarchical clustering is used to subdivide the user groups based on the collected user data. This process calculates the similarity between different users and classifies users with similar behavior characteristics into a category, thereby identifying several subgroups with typical behavior patterns. For example, one group of users may exhibit the characteristics of high-frequency browsing but low conversion, while another group of users may prefer high-priced rentals with longer rental periods.

[0066] This user clustering method based on clustering algorithms can automatically discover structured features in user groups without the need for manual rule definition, effectively improving the scientificity and objectivity of user classification. At the same time, it also provides a basis for the differentiated design of subsequent recommendation strategies, enabling the platform to develop more targeted service solutions based on different user groups.

[0067] Finally, based on the completion of user clustering, the system further combines the overall characteristics of each user's group and the individual's historical behavior trajectory to construct a user portrait containing multiple dimensional labels such as interest preferences, consumption capacity, and geographic location preferences. This portrait not only reflects the user's current interest points but also predicts the types of rental products that may interest them in the future.

[0068] By constructing a refined user portrait system, the platform can achieve a transition from "extensive recommendation" to "personalized recommendation", significantly improving the relevance and matching degree of recommended content. In addition, the design of multi-dimensional labels also enhances the explainability of the recommendation system, making the recommendation results more transparent and credible, which helps to improve user trust and satisfaction.

[0069] In summary, this implementation method integrates user basic information and behavior data, uses clustering analysis and multi-dimensional portrait technology to achieve fine division and in-depth insight into user groups, providing a solid data foundation and technical support for personalized recommendation in the rental business scenario.

[0070] In some embodiments of the present application, a target algorithm model is selected and trained according to the user portrait and the characteristics of the rental product, and the model parameters are optimized through cross-validation, specifically:

[0071] Based on the multi-dimensional labels in the user portrait and the attribute information of the rental product, a user-item interaction feature matrix is constructed;

[0072] In combination with collaborative filtering, content-based recommendation or hybrid recommendation algorithm, a target recommendation model suitable for the leasing business scenario is selected;

[0073] The model is parameter-optimized using a cross-validation method, and the model performance is measured by accuracy, recall rate and AUC value to ensure the accuracy of the recommendation results.

[0074] In this embodiment, based on the constructed user portrait and leasing product attribute information, the system further generates a user-item interaction feature matrix. This matrix associates the multi-dimensional labels in the user portrait (such as interest preferences, consumption ability, geographical location inclination, etc.) with the key features of the leasing products (such as rent, lease period, product category, equipment status, etc.), forming structured input data for the training of the recommendation model.

[0075] By constructing the user-item interaction feature matrix, the system can convert complex user behavior and product attributes into numerical expressions understandable by machine learning models. This structured representation not only improves the model's understanding of user preferences, but also provides a high-quality data foundation for subsequent algorithm training, thereby enhancing the intelligence level and prediction accuracy of the recommendation system.

[0076] Subsequently, the system selects a target recommendation model that best fits the leasing business scenario by combining collaborative filtering, content-based recommendation or hybrid recommendation algorithm. Collaborative filtering algorithm can mine the similarity between users or the correlation between items, which is suitable for scenarios with rich interaction data; content-based recommendation focuses on analyzing the attributes of leasing products, which is suitable for cold start or new user / new product recommendation; and hybrid recommendation combines the advantages of both, improving overall recommendation performance.

[0077] By flexibly selecting different types of recommendation algorithms and combining them according to actual business needs, the system can achieve a good balance in recommendation accuracy, coverage, diversity, etc. This adaptive design makes the recommendation model more suitable for the characteristics of the leasing industry, such as long lease period, complex decision-making, clear user intent, etc., thereby improving the effectiveness and practicality of the recommendation.

[0078] Finally, during the model training process, the system uses a cross-validation method to optimize the model parameters, and measures the model performance through evaluation indicators such as precision (Precision), recall rate (Recall) and AUC value. Cross-validation effectively prevents model overfitting by dividing and validating the training data multiple times, improving the generalization ability; and the accurate evaluation system provides a clear direction for model iteration and optimization.

[0079] This mechanism significantly enhances the stability and robustness of the recommendation model, ensuring that it maintains good recommendation quality even when faced with large-scale user groups and dynamically changing market environments. At the same time, by continuously monitoring model performance and regularly updating training data, the system has the ability to evolve itself, adapting to changing user needs and business trends.

[0080] In summary, this implementation builds a user-item interaction matrix, selects a recommendation algorithm that fits the leasing business, and introduces a scientific model evaluation and optimization mechanism, forming an efficient, stable, and scalable personalized recommendation system. This system not only improves the accuracy of recommendation results but also enhances the intelligent service capabilities of the platform, helping to improve user satisfaction and business conversion efficiency.

[0081] In some embodiments of the present application, the recommendation results are dynamically adjusted according to the behavior data of the current session when the user accesses, and multi-channel pushing is implemented, specifically:

[0082] Real-time collection of user behavior data in the current session, including search keywords, page browsing, dwell time, and click preferences;

[0083] Lightweight updating of the recommendation model based on online learning mechanism, dynamically adjusting the recommendation list to match the user's immediate interest;

[0084] Push the updated recommendation results through multiple channels, including platform homepage recommendation positions, emails, mobile application notifications, and SMS.

[0085] In this embodiment, the system collects the user's behavior data in the current session in real time during the user's access to the platform, including search keywords, page browsing path, page dwell time, and click preferences, etc. These behavior data can reflect the user's immediate interest and potential demand in the current access stage, and are an important basis for dynamically adjusting the recommended content.

[0086] Compared with the static user portrait constructed only by relying on historical behavior, the introduction of real-time session behavior analysis can more accurately capture the short-term intention changes of users. For example, a user who usually prefers to rent office equipment may indicate a new temporary demand if they frequently browse outdoor equipment pages during this access. By capturing these subtle behavior changes, the system can adjust the recommendation strategy in a timely manner, improving the relevance and timeliness of the recommendations.

[0087] Next, the system updates the recommendation model based on an online learning mechanism. "Online learning" refers to the model's ability to quickly adjust the recommendation results using the latest user behavior without retraining the entire model. This mechanism is particularly suitable for the rental business, where user intentions are variable and rental periods are flexible, allowing the recommendation system to respond in real-time with low computational cost.

[0088] The lightweight online learning update strategy not only improves the real-time adaptability of the recommendation system, but also effectively reduces resource consumption and system latency, ensuring stable operation in high-concurrency scenarios. This provides scalable technical support for large-scale rental platforms, enabling personalized recommendations to be both efficient and meet user experience needs.

[0089] Finally, the system pushes the updated recommendation results through various channels, including platform homepage recommendation positions, emails, mobile application notifications, and SMS. This multi-channel reach mechanism ensures that users receive highly relevant rental product information whether they are actively accessing the platform or offline.

[0090] The multi-channel push strategy not only enhances the coverage of recommended content but also significantly improves user reach and interaction conversion rates. In the rental business, users often require multiple touchpoints and guidance to make decisions, so providing relevant recommendations through diverse push forms helps improve user stickiness, promotes transaction completion, and enhances overall platform operational efficiency.

[0091] In summary, this implementation introduces real-time behavior collection, online learning mechanisms, and multi-channel push strategies, upgrading the recommendation system from "static recommendation" to "dynamic response" and greatly improving recommendation flexibility and accuracy. It not only improves user experience but also brings higher conversion rates and operational efficiency to rental platforms.

[0092] In some embodiments of the present application, the method further comprises:

[0093] A user feedback mechanism is introduced to allow users to rate, like, collect, or mark as uninterested in the recommended results.

[0094] Based on user feedback data, negative and positive samples are constructed to optimize the training and iteration of the recommendation model.

[0095] In this embodiment, the system further introduces a user feedback mechanism, allowing users to actively evaluate the recommended results, including rating, liking, collecting, or marking as "uninterested." These feedback behaviors constitute users' true attitude towards recommended content and are one of the important bases for measuring recommendation quality.

[0096] Compared with the method of relying only on implicit behaviors (such as clicks, browsing) for model training, the explicit feedback mechanism can more directly obtain the user's preference for the recommendation results, thereby improving the understanding ability and response speed of the recommendation system. By incorporating the user's subjective evaluation into the data system, the system can more accurately identify whether the recommended content meets the user's actual needs, providing high-quality training signals for subsequent model optimization.

[0097] Based on the collected user feedback data, the system further constructs positive and negative samples for the training and iterative optimization of the recommendation model. Specifically, the user's likes, collections, or high ratings are defined as positive samples, indicating the user's approval of the recommended content; and the behaviors of marking "not interested" or low ratings are used as negative samples, reflecting the deviation of the recommended content from the user's interests.

[0098] This sample construction method based on user feedback not only enriches the diversity of training data, but also effectively alleviates the "exposure bias" and "cold start" problems existing in traditional recommendation systems. By continuously accumulating high-quality labeled samples, the recommendation model can better learn the user preference patterns during the training process, improving the accuracy and personalization level of the recommended results.

[0099] In summary, this implementation introduces a user feedback mechanism and constructs positive and negative samples based on feedback data, realizing the transition of the recommendation system from "passive recommendation" to "active learning". This mechanism enhances the system's understanding of the user's true intentions, improves the training efficiency and generalization performance of the recommendation model,

[0100] In some embodiments of the present application, the method further comprises:

[0101] Periodically synchronizing and updating user behavior data and rental product information from the platform database to ensure the timeliness and completeness of the data relied on by the recommendation system;

[0102] Weighted processing of historical behavior data based on a time decay function.

[0103] In this embodiment, the system further introduces a periodic data synchronization mechanism, that is, according to a preset time period (such as daily, weekly, or real-time streaming update), the latest user behavior data and rental product information are synchronized from the platform database. This mechanism ensures that the recommendation system is always based on the latest data for modeling and prediction, avoiding the deviation of the recommended results from the user's true interests due to data lag.

[0104] User behavior data (such as browsing, clicking, saving, and ordering) and rental product information (such as inventory status, price adjustments, and new categories) can change significantly over time. If a recommendation system relies on static data for a long period, the model will become outdated, reducing recommendation accuracy and user experience. By regularly synchronizing and updating, the system can promptly capture shifts in user interests and market dynamics, thereby maintaining the relevance and timeliness of recommended content.

[0105] Furthermore, the system employs a historical behavior weighting method based on a time decay function to dynamically adjust the weights of users' historical behavior data. Specifically, behavior data closer to the current time is assigned higher weights, while earlier historical behaviors are gradually reduced in influence using an exponential or linear decay method to reflect the evolving trends of user interests.

[0106] This time decay mechanism effectively solves the problem of "historical preferences dominating future recommendations" in traditional recommendation systems, making the model pay more attention to users' recent behavior and improving the agility and accuracy of personalized recommendations. For example, if a user previously preferred renting photography equipment but has recently been frequently browsing office equipment pages, the system should gradually reduce the recommendation weight for camera products and increase the recommendation weight for office equipment such as laptops.

[0107] In summary, this implementation method constructs a recommendation system with self-updating and time-aware capabilities by periodically synchronizing and updating the data source and introducing a time decay function to weight historical behavior. This not only improves the response speed and adaptability of the recommendation system but also enhances the model's accuracy in understanding the evolution of user interests.

[0108] In some embodiments of this application, the method further includes:

[0109] Analyze users' historical behavior and real-time interaction data to identify potential abnormal operations or malicious order-brushing behavior;

[0110] Using techniques such as graph neural networks or isolated forest algorithms, user behavior graphs are constructed and suspicious nodes are identified.

[0111] In this embodiment, the system further introduces a user behavior anomaly detection mechanism. By comprehensively analyzing users' historical behavior and real-time interaction data, it identifies potential abnormal operations or malicious order-brushing behaviors. These behaviors may include a large number of fake clicks in a short period of time, abnormal order placement, frequent order cancellations, abnormal browsing paths, etc., which seriously interfere with the accuracy of the recommendation system and affect the platform's operational security.

[0112] Traditional recommendation systems often focus only on modeling user interests, while ignoring the legitimacy and authenticity of user behavior. By introducing a behavior anomaly recognition mechanism, the system can assess the risk of user behavior before making recommendations, filtering out unreliable data sources and improving the credibility of the recommendation results and the platform's data governance capabilities.

[0113] To more efficiently identify abnormal behavior in complex patterns, the system uses advanced techniques such as graph neural networks (GNN) or isolation forest algorithms to construct a user behavior graph and identify suspicious nodes. Graph neural networks can model the interaction between users and items as a graph structure, uncovering patterns of association between groups, while isolation forests are good at finding data points that deviate from the normal distribution in an unsupervised environment, making them suitable for high-dimensional sparse behavior data identification.

[0114] Through the combination of graph models and anomaly detection algorithms, the system can not only identify individual user behavior, but also discover potential collusion groups such as single brushing alliances or multi-account manipulation behavior. This joint analysis method based on graph structure and statistical learning significantly improves the comprehensiveness and accuracy of anomaly identification, enhancing the platform's risk prevention and control capabilities.

[0115] In summary, this implementation method achieves quality control of the input data of the recommendation system by constructing a user behavior graph and applying advanced anomaly detection algorithms, effectively improving the safety and reliability of the recommended content.

[0116] As shown in Figure 2 the second aspect embodiment of the present application provides a rental business personalized data recommendation device, which comprises:

[0117] The classification module 110 is adapted to extract the basic information and behavior data of the user from the platform database, subdivide the user groups based on the basic information and behavior data of the user using a clustering algorithm, and establish a detailed user portrait for each user;

[0118] The parameter optimization module 120 is adapted to select a target algorithm model according to the user portrait and the characteristics of the rental product and train it, and optimize the model parameters through cross-validation;

[0119] The push module 130 is adapted to dynamically adjust the recommendation results according to the behavior data of the current session when the user accesses, and implement multi-channel push.

[0120] The rental business personalized data recommendation device provided by the second aspect embodiment of the present application can implement the rental business personalized data recommendation method in any of the above first aspect embodiments, and therefore can achieve any of the technical effects of the above rental business personalized data recommendation method, which will not be described here.

[0121] The third aspect of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method for recommending personalized data of rental business in any of the embodiments of the first aspect.

[0122] Figure 3 An example of a schematic diagram of the physical structure of an electronic device is shown in FIG. 8. Figure 3 As shown in FIG. 8, the electronic device can include a processor 810, a communication interface 820, a memory 830, and a communication bus 840, wherein the processor 810, the communication interface 820, and the memory 830 can communicate with each other through the communication bus 840. The processor 810 can invoke the logical instructions in the memory 830 to execute the method for recommending personalized data of rental business in any of the embodiments of the first aspect, which comprises the following steps:

[0123] Step 100: extracting basic information and behavior data of a user from a platform database, subdividing a user group based on the basic information and behavior data of the user using a clustering algorithm, and establishing a detailed user portrait for each user.

[0124] Step 200: selecting a target algorithm model according to the user portrait and the characteristics of rental products and training the target algorithm model, and optimizing model parameters through cross-validation.

[0125] Step 300: dynamically adjusting the recommendation result according to the behavior data of a current session when the user accesses, and implementing multi-channel pushing.

[0126] In addition, the logical instructions in the memory 830 described above can be implemented in the form of a software functional unit and sold or used as an independent product, which can be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the present application or parts of the present application that essentially contribute to the prior art or the parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in the embodiments of the present application. The aforementioned storage medium includes a U disk, a mobile hard disk, a read-only memory, a random access memory, a magnetic disk or an optical disk, and various media that can store program codes.

[0127] On the other hand, the present application also provides a computer program product, which comprises a computer program, the computer program can be stored on a non-transitory computer readable storage medium, and the computer program is executed by a processor, so that the computer can execute the method for recommending personalized data of rental business provided by the above-mentioned methods, which comprises the following steps:

[0128] Step 100, extract the basic information and behavior data of the user from the platform database, subdivide the user group based on the basic information and behavior data of the user using a clustering algorithm, and establish a detailed user portrait for each user.

[0129] Step 200, select a target algorithm model according to the user portrait and the characteristics of the rental product and train it, and optimize the model parameters through cross-validation.

[0130] Step 300, dynamically adjust the recommendation result according to the behavior data of the current session when the user accesses, and realize multi-channel push.

[0131] In another aspect, the present application also provides a non-transitory computer readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the rental business personalized data recommendation method provided by the above-mentioned methods, and the method comprises:

[0132] Step 100, extract the basic information and behavior data of the user from the platform database, subdivide the user group based on the basic information and behavior data of the user using a clustering algorithm, and establish a detailed user portrait for each user.

[0133] Step 200, select a target algorithm model according to the user portrait and the characteristics of the rental product and train it, and optimize the model parameters through cross-validation.

[0134] Step 300, dynamically adjust the recommendation result according to the behavior data of the current session when the user accesses, and realize multi-channel push.

[0135] Finally, the present application also provides a non-volatile computer storage medium having computer executable instructions stored thereon, which, when executed by a processor, implements the rental business personalized data recommendation method provided by the above-mentioned methods, and the method comprises:

[0136] Step 100, extract the basic information and behavior data of the user from the platform database, subdivide the user group based on the basic information and behavior data of the user using a clustering algorithm, and establish a detailed user portrait for each user.

[0137] Step 200, select a target algorithm model according to the user portrait and the characteristics of the rental product and train it, and optimize the model parameters through cross-validation.

[0138] Step 300, dynamically adjust the recommendation result according to the behavior data of the current session when the user accesses, and realize multi-channel push.

[0139] The places not mentioned in the present application can be realized by adopting or referring to the existing technology.

[0140] The various embodiments in the specification are described in progressive manner, and the same or similar parts between the various embodiments can be mutually referred to, and each embodiment focuses on the difference from other embodiments.

[0141] The above merely provides an example of the present application, and is not intended to limit the present application. The present application can have various changes and modifications for those skilled in the art. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A personalized data recommendation method for leasing business, characterized in that, include: Extract basic user information and behavioral data from the platform database, use clustering algorithms to segment user groups based on the basic user information and behavioral data, and create a detailed user profile for each user; Based on the user profile and rental product characteristics, a target algorithm model is selected and trained, and the model parameters are optimized through cross-validation. The system dynamically adjusts recommendation results based on the user's behavior data during the current session and enables multi-channel push notifications.

2. The personalized data recommendation method for leasing business according to claim 1, characterized in that, The process involves extracting basic user information and behavioral data from the platform database, using clustering algorithms to segment user groups based on this information and data, and creating a detailed user profile for each user. Specifically: Collect users' basic attribute information and historical behavior data, including age, gender, occupation, browsing history, click rate, favorites and order records; Users are grouped using K-means or hierarchical clustering algorithms to identify subgroups of users with similar behavioral patterns and preferences; By combining the characteristics of the user's group and individual behavior patterns, a multi-dimensional user profile is constructed, including tags such as interests, spending power, and geographical location preferences.

3. The personalized data recommendation method for leasing business according to claim 2, characterized in that, The process of selecting and training a target algorithm model based on the user profile and rental product characteristics, and optimizing the model parameters through cross-validation, specifically involves: Based on the multi-dimensional tags in user profiles and the attribute information of rental products, a user-item interaction feature matrix is ​​constructed. By combining collaborative filtering, content-based recommendation, or hybrid recommendation algorithms, a target recommendation model suitable for the rental business scenario is selected. Cross-validation was used to fine-tune the model parameters, and the model performance was measured by accuracy, recall, and AUC to ensure the accuracy of the recommendation results.

4. The personalized data recommendation method for leasing business according to claim 1, characterized in that, The method of dynamically adjusting recommendation results based on the user's behavior data during the current session and implementing multi-channel push notifications specifically includes: Real-time collection of user behavior data in the current session, including search keywords, pages viewed, dwell time, and click preferences; The recommendation model is updated in a lightweight manner based on an online learning mechanism, and the recommendation list is dynamically adjusted to match the user's real-time interests. The updated recommendation results will be pushed through multiple channels, including homepage recommendations, emails, mobile app notifications, and SMS.

5. The personalized data recommendation method for leasing business according to any one of claims 1 to 4, characterized in that, The method also includes: Introduce a user feedback mechanism, allowing users to rate, like, favorite, or mark as uninterested in the recommendation results; Negative and positive samples are constructed based on user feedback data to optimize the training and iteration of the recommendation model.

6. The personalized data recommendation method for leasing business according to any one of claims 1 to 4, characterized in that, The method also includes: Regularly synchronize and update user behavior data and rental product information from the platform database to ensure the timeliness and completeness of the data on which the recommendation system relies; Historical behavior data is weighted based on a time decay function.

7. The personalized data recommendation method for leasing business according to any one of claims 1 to 4, characterized in that, The method also includes: Analyze users' historical behavior and real-time interaction data to identify potential abnormal operations or malicious order-brushing behavior; Using techniques such as graph neural networks or isolated forest algorithms, user behavior graphs are constructed and suspicious nodes are identified.

8. A personalized data recommendation device for rental business, characterized in that, include: The classification module is suitable for extracting basic user information and behavioral data from the platform database, using clustering algorithms to segment user groups based on the basic user information and behavioral data, and creating a detailed user profile for each user. The parameter optimization module is adapted to select a target algorithm model and train it based on the user profile and rental product characteristics, and optimize the model parameters through cross-validation. The push module is suitable for dynamically adjusting recommendation results based on the user's behavior data during the current session and enabling multi-channel push.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the personalized data recommendation method for leasing business as described in any one of claims 1 to 7.

10. A non-volatile computer storage medium storing computer-executable instructions thereon, characterized in that, When the computer program is executed by the processor, it implements the personalized data recommendation method for leasing business as described in any one of claims 1 to 7.

Images