DYNAMIC LIFE SPAN DETERMINATION SYSTEM FOR BASKET DATA

Abstract

This invention relates to a system (1) developed to enable the implementation of a dynamic lifecycle policy specific to each cart item by processing the user's past purchase dynamics, real-time on-site behavioral click data and real-time stock risk scores of the items added to the cart with machine learning models.

Description

1 TARIFF DYNAMIC LIFE SPAN DETERMINATION SYSTEM FOR BASKET DATA Technical Area This invention analyzes a user's past purchasing dynamics in real-time on-site behavior. click data and instant stock risk scores of products added to the cart are machined. By processing it with learning models, each basket object has a unique, dynamic life. It relates to a system that enables the implementation of a cycle policy. Previous Technique In current e-commerce systems, products with limited stock or high demand products being added to the cart by users with low purchase intention, products remaining reserved for a long time without actually being used and the actual stock level being misleading 15 This can cause it to appear in this way. In addition, millions of active Full Object JSON (JSON) of shopping cart information on platforms with user interfaces. Keeping it in memory for a long time in this form leads to unnecessary RAM consumption and related problems. This also leads to high server and cloud hosting costs. Furthermore... If the baskets are automatically deleted after a certain period, deletion 20 the decision depends on user behavior, shopping cart abandonment rate, repurchase probability and related factors. The financial effects are not systematically measured, and the results obtained are not used in subsequent baskets. The decisions to keep and delete are not fed back into the system. Therefore, current solutions... Efficient management of inventory usage, dynamic use of memory and infrastructure resources. 25 It is inadequate in terms of being managed through this mechanism. 2 Therefore, the user's past purchase dynamics and current on-site behavior... click data and instant stock risk scores of products added to the cart are machined. By processing it with learning models, each basket object has a unique, dynamic life. a system is needed that enables the implementation of the cycle policy It is understood. 5 According to Chinese patent document number CN111144987, which is included in the prior art, Methods of limiting abnormal shopping behavior, limitation mechanisms, and shopping. The system in question is being discussed. This invention addresses abnormal shopping behaviors. The restriction method, the mechanism for restricting abnormal shopping behavior, and a shopping 10 The system describes the method; user access to a shopping cart interface. the collection of data; user access data to a big data platform sending and using an identification model of the big data platform Identifying abnormal users based on user access data; Obtaining user information belonging to the abnormal user and the user in question 15 the information is stored on a cache server and in a predetermined manner. According to the restriction rule, an access limit is set for the abnormal user. This includes reducing the shopping cart load caused by the user. a mechanism to limit abnormal shopping behavior using the restriction method Thanks to this, abnormal shopping behaviors such as automatic order replenishment are prevented. creating a load on the shopping cart system and hindering other users' normal shopping experience. Problems affecting the experience are being resolved. The shopping system, abnormal shopping It includes a mechanism to limit the behavior of abnormal users and their orders. shopping cart caused by constant replenishment or similar shopping behaviors It reduces the burden and improves the users' shopping experience. 25 Brief Description of the Invention 3 The aim of this invention is to analyze a user's past purchase dynamics in real-time on the site. behavioral click data and real-time stock risk scores of products added to the cart processed with machine learning models, a dynamic, unique design is created for each basket object. a system developed to ensure the implementation of life cycle policy to accomplish. 5 Detailed Description of the Invention The “Dynamic Lifecycle for Basket Data” system was developed to achieve the purpose of this invention. The "Time Determination System" is shown in the attached figure; this figure is 10. Figure 1 shows a schematic view of the system that is the subject of the invention. The parts shown in the figure are individually numbered, and the corresponding numbers correspond to these numbers. given below. 15 1. System 2. Application 3. Database 4. Server 20 User's past purchase dynamics, real-time on-site behavioral clicks. machine learning of data and real-time stock risk scores of products added to the cart processed with models, each basket object has a unique dynamic lifecycle. The system in question, developed to ensure the implementation of the policy (1); 25 - configured to enable the user to carry out e-commerce transactions at least one application (2), 4 -order data, user data, product data, AI outputs, and at least one structured to ensure that reporting data is kept on record database (3), -the probability of the user's (2) shopping cart in the application being completed with a purchase predicting, generating a purchase tendency score, and 5 items in the basket the risk it creates in terms of inventory blockage, opportunity cost and supply pressure calculation, purchasing tendency, inventory risk, system resource utilization and basket combining multiple signals in context into a single decision state vector, selecting the most appropriate lifecycle action for the basket, lifecycle policy 10. Defining the outputs and actually applying the action on the basket object. It includes at least one server (4) configured to provide. The application (2) in the system (1) which is the subject of the invention does not require any communication protocol. to communicate with the database (3) and server (4) and exchange data using It is structured to accomplish this. 15 The database (3) in the system (1) which is the subject of the invention does not use any communication protocol. to communicate and exchange data with the application (2) and the server (4) using It is structured to accomplish this. The server (4) in the system (1) which is the subject of the invention, any communication protocol to communicate and exchange data with the application (2) and the database (3) using It is configured to perform. The server (4) has a fixed value for each basket object. Instead of applying a TTL (Time-To-Live) value, basket-based contextual scores are used. the production and complete storage of the basket according to these scores, abbreviated 25 the decision to retain it for a period of time, convert it to a simplified format or delete it It is configured to enable the user to add products to the cart. The server (4) adding, updating the cart, viewing the cart, or completing the purchase process Identifying instances where the session is terminated before completion, for each instance subsequently, the user's past order frequency and conversion rate for the relevant shopping cart item. and average basket value, in-session browsing behavior, product review time, basket Addition order and historical cart abandonment rate, stock quantity of items in the cart, inventory turnover rate, margin information and campaign status, cache fill rate, cart item size, and active 5 Creating a feature set by defining parameters such as basket density. It is configured to provide gradient boosting based on the server (4). Classification algorithms use features collected to determine the purchase order of the relevant basket. calculating the probability of this outcome, the purchase tendency score of the model output. 10. This ensures that the basket represents the likelihood of conversion in the short term. It is configured accordingly. The server (4) displays the products in the basket on the stock. The impact should also be evaluated, including the remaining stock quantity of the product and the specific timeframe. parameters such as sales speed, delivery time and campaign criticality within the range The stock risk score is calculated using this method, and the score reflects the long retention period of the relevant product in the shopping cart. 15 It is structured to provide the server (4), purchase tendency score, stock risk A decision is made for the basket by evaluating the score and feature parameters together. the vector is generated and the vector purchase tendency score, stock risk score, user the likelihood of revisiting the cart and the time elapsed since the last interaction It is configured to include variables. Server (4), 20 the generated state vector is processed by a decision engine, decision contextual multi-armed bandit, reinforcing Reinforcement learning, or a rule engine that works in conjunction with them. The decision engine's ability to have this structure, for each basket object, is to determine the complete object of the basket. Storing as standard TTL, the basket as full object abbreviated TTL 25 storage, converting the basket to a lighter basket format (Lite Cart), the basket retaining the semantic representation for the purpose of deletion and retrieval, and lowering the basket. one of the actions to move to a costly secondary storage layer 6 instead of applying a uniform timeout for each basket, the commercial value of the basket should be considered. to ensure that both technical and technical costs are optimized It is configured. The server (4) converts the decision engine to a simplified basket format. If the user chooses the conversion action, the basket object will have a lightweight data version. conversion to its current structure, including product images, long description fields, 5 during the conversion process heavy data fields for campaign components and user interface removal of only the minimum data necessary for reconstruction the retention of the minimum data structure in question, preferably user ID, product It should include the identity, quantity information, time added to cart, and basic price information, and the promise. The subject of the transaction is 10% cache consumption without completely losing the context of the shopping cart. It is configured to ensure that the basket is reduced. Server (4), where the basket is lightened or, in cases where it is deleted, the contents of the shopping cart can be embedded into a semantic vector using an embedding model. its transformation into a representation, the subsequent reconstruction of the basket context, and The generated vector includes the category, price range, brand, and usage of the products in the shopping cart. It is a compressed representation of the context and records 15 in the vector database (3). It is configured to ensure that the user is kept under control. The server (4) ensures that the user is kept under control. If the full basket object is not available when the application (2) returns using a simplified basket format data and / or vector representation of the basket recreating the cart if one of the items in the cart is out of stock, Vector similarity search and rule engine help you find similar categories, price ranges, or 20s. Restructuring the shopping cart by identifying alternative products with similar product features. It is configured to provide lifecycle decisions. The server (4) updating the decision-making policy by monitoring the results, after the basket is deleted whether the user is searching for the same product again, reduced cart return whether or not it was done, whether the purchase took place, and how much 25 was involved with the implemented decision. parameters such as whether memory savings are achieved and whether inventory risk is reduced measuring the outputs, using the outputs as a reward or feedback function, periodic or continuous retraining of the decision engine and user 7 the balance between behaviors and infrastructure resource consumption over time To ensure that appropriate lifecycle policies are developed for each basket through learning and analysis. It is structured accordingly. Industrial Applicability 5 Thanks to the system (1) that is the subject of the invention, the user's past purchase dynamics can be instantly accessed. On-site behavioral click data and real-time stock risk of products added to the cart. The scores are processed by machine learning models to create unique scores for each basket object. This ensures the implementation of a dynamic life cycle policy. 10 Around these fundamental concepts, the invention topic is "Dynamic Lifecycles for Basket Data". It is possible to develop a wide variety of applications related to the Time Determination System (1)”. and the invention cannot be limited to the examples described here, but mainly to the claims as stated. 15

Claims

8 REQUESTS 1. User's past purchase dynamics, real-time on-site behavioral clicks. the machine's real-time stock risk scores of the data and products added to the cart By processing it with learning models, a dynamic 5 unique to each basket object is created. enabling the implementation of life cycle policy; -to enable the user to carry out e-commerce transactions containing at least one configured application (2) and -order data, user data, product data, artificial intelligence 10 to ensure that outputs and reporting data are kept on record at least one structured database (3), -completion of the user's (2) shopping cart in the application with purchase estimating the probability, generating a purchase propensity score, and including items in the basket. In terms of stock blocking of the available products, opportunity cost and supply pressure calculating the risk it creates, purchasing tendency, inventory risk, system resources 15 Multiple signals in the form of usage and basket context form a single decision situation. combining it in the vector, the most suitable lifecycle action for the basket selection, definition of life cycle policy outcomes and action the most configured to actually implement it on the basket object a system characterized by a small number of servers (4) (1). 20 2. Using any communication protocol, connect the database (3) and the server (4) structured for communication and data exchange A system like the one in Claim 1 characterized by application (2) (1).

3. Using any communication protocol, the application (2) and the server (4) data structured for communication and data exchange A system like the one in Claim 1 or 2 characterized by its base (3) (1). 9 4. Using any communication protocol, application (2) and database (3) structured to communicate and exchange data as in any of the above requests characterized by the server (4) a system (1). 5 5. Applying a fixed TTL (Time-To-Live) value for each basket object. Instead, generating basket-based contextual scores and acting according to these scores storing the basket properly, keeping it for a short period of time, lightened 10 to enable a decision to be made on whether to convert to a different format or delete it any of the above requests characterized by the configured server (4) a system like one of them (1).

6. User adding products to the cart, updating the cart, viewing the cart or ending the session before completing the purchase 15 the detection of events, and after each event, the relevant basket object user's past order frequency, conversion rate, and average basket value, in-session navigation behavior, product review time, cart order, and historical cart abandonment rate, number of items in the cart in stock, inventory turnover rate, margin information and campaign status, cache fill rate, cart item size and 20 By defining parameters such as active basket density, a feature set is created. characterized by the server (4) configured to enable its creation a system like any of the above requests (1).

7. Features collected using gradient boosting-based classification algorithms 25 using the probability of the relevant basket resulting in a purchase the calculation, the model output being the purchase tendency score, and the basket to ensure that it represents the possibility of transformation in the short term any of the above requests characterized by the configured server (4) a system like one of them (1).

8. The impact of the products in the basket on inventory should also be evaluated. The remaining stock quantity of the product, the sales rate within a specific time period, and the delivery time are all factors considered. The stock risk score is calculated using parameters such as campaign criticality. the calculation and score of the amount of stock that the relevant product has kept in the cart for a long time to ensure it expresses the risk of creating a blockage or opportunity cost any of the above requests characterized by the configured server (4) a system like one of them (1). 10 9. Purchase propensity score, stock risk score, and feature parameters together. by evaluating the basket, a decision state vector is created for the vector. purchase propensity score, stock risk score, user's cart revisit rate Variables such as probability and time elapsed since the last interaction 15 characterized by the server (4) configured to enable it to contain a system like any of the above requests (1).

10. Processing the generated state vector by a decision engine, decision The contextual multi-armed bandit, 20 Reinforcement learning or learning that works in conjunction with them It can be in the structure of a rule engine, each basket object of the decision engine For this, storing the basket as a complete object with standard TTL, the basket as a complete object Storing items in TTL format, abbreviated as TTL, in a simplified cart format. (Lite Cart) conversion, deletion of cart and retrieval purposes semantically 25 maintaining representation and a lower-cost secondary storage of the basket choose one of the actions such as moving to the next layer and for each basket Instead of applying a uniform time limit, the commercial value and technical aspects of the basket should be considered. server configured to ensure cost optimization together 11 (4) like any of the above-mentioned claims characterized by system (1).

11. Converting the decision engine's basket to a lighter basket format. If the user chooses this action, the shopping cart object will have a lightweight data structure. conversion, product images, long description fields during conversion, heavy data fields for campaign components and user interface removal of only the minimum data necessary for reconstruction the retention of the minimum data structure in question, preferably the user ID, Product ID, quantity information, cart addition time, and basic price information. 10 and without completely losing sight of the context of the transaction and the basket. The server is configured to ensure reduced cache consumption (4) a system like any of the above characterized demands (1).

12. In cases where the cart is lightened or deleted, the cart contents will be embedded. 15 converting the basket context into a semantic vector representation using the model, then it can be reinstalled and the generated vector can be categorized by product in the basket. a compressed representation that illustrates the price range, brand, and usage context to ensure that it is kept under record in the vector database (3) Any of the above requests characterized by the configured server (4) 20 a system like one of them (1).

13. Full cart object when the user returns to the application (2) if not available, a simplified basket format data and / or vector representation using the method of recreating the cart, if there are 25 items in the cart If one is out of stock, vector similarity search and rule engine assistance can be used. alternative products in a similar category, price range, or product feature configured to enable basket restructuring by determining 12 as in any of the above requests characterized by the server (4) a system (1).

14. Monitoring the outcomes of life cycle decisions to improve decision-making policy. the update, after the cart is deleted, the user searches for the same product again and 5 whether he / she called, whether he / she returned the reduced basket, purchase whether it happened or not, and how much memory is saved with the implemented decision. parameters such as whether it is provided and whether the inventory risk is reduced measuring outputs as a reward or feedback function the use of the decision engine, periodic or continuous re-use 10 training and correlation between user behavior and infrastructure resource consumption learning the balance over time and finding the appropriate life cycle for each basket server (4) configured to enable the production of policies a system like any of the above characterized demands (1).

Images