A smart product recommendation method

Abstract

E-commerce has become a mainstream lifestyle. Users primarily rely on platform-provided content recommendation algorithms to find the goods or services they need. Recommendations are generated based on VR / AR data (CN118552273A) using virtual digital avatars, and inferred from user social connections (CN118411232A). Existing recommendation algorithms lack personalization, over-rely on user behavior and sales data, and are susceptible to data pollution from fraudulent activities like fake orders, leading to results that don't meet user expectations, causing confusion, and ultimately affecting user decision-making and purchasing decisions. To address these issues, this application provides an intelligent method for recommending goods and services. It innovatively modifies existing recommendation algorithms by introducing advanced technologies such as large-scale artificial intelligence models and computer image recognition algorithms. This method effectively improves the personalization of user search results, creating scenario-specific, personalized search results. This allows search results to help users make faster decisions and increase their willingness to purchase products and services.

Description

Technical Field

[0001] This invention relates to the fields of e-commerce and artificial intelligence, and specifically to a method for recommending intelligent goods and services. Background Technology

[0002] E-commerce has become a mainstream lifestyle. Users primarily rely on platform-provided content recommendation algorithms to find the goods or services they need when shopping. This includes recommendation data generated from VR / AR data based on virtual digital avatars (CN118552273A) and recommendation data inferred based on user social connections (CN118411232A).

[0003] Existing recommendation algorithms lack personalized features, and search results rely excessively on user behavior and sales data. The recommendation results are easily polluted by data pollution caused by behaviors such as fraudulent orders, resulting in recommendations that do not meet user expectations, causing confusion for users, and thus affecting users' decision-making and willingness to place orders. Summary of the Invention

[0004] To address the aforementioned issues, this application provides an intelligent product and service recommendation method. It innovatively modifies existing recommendation algorithms by incorporating advanced technologies such as large-scale artificial intelligence models and computer image recognition algorithms. This method effectively enhances the personalization of user search results, creating scenario-specific results tailored to each individual. This allows search results to help users make faster decisions and increase their willingness to purchase products and services. Method implementation steps (see) Figure 1 )

[0005] Step 1: The user authorizes the platform to collect the user's personal image data through hardware or user uploads. This step fully considers the protection of the user's privacy data, preventing the acquisition of the user's biometric information and facial features without authorization. If the user does not authorize the platform to collect video data, the user can provide this information through text description.

[0006] Step 2: The platform obtains the user's body and facial features through authorized hardware or file uploads, scanning, etc., to facilitate subsequent analysis of the user's appearance and physical characteristics.

[0007] Step 3: The system performs a usability test on the collected content to ensure it contains compliant image information. This includes, for example, a complete human body or upper body information, a single human figure occupying at least 50% of the image, and facial color and saturation meeting the analysis criteria. Data that does not meet these criteria will prompt the user to re-upload.

[0008] Step 4: Image background separation. Data in the image that is not involved in subsequent analysis is separated from the background using automatic or manual annotation methods.

[0009] Step 5: Enter or select scene information. Users specify the scene information in which the image should appear by selecting or manually entering it. This step aims to make the system's suggestions more scene-appropriate. Step 6: Use a large-scale artificial intelligence model to organize and analyze the information from Steps 4 and 5. In this step, the platform will continuously optimize and adjust specific prompt information to generate analysis results that are more suitable for the user's specific scenario. Step 7: The platform automatically extracts the feature keywords generated in Step 6 and uses them as search criteria to search for products and services on the platform or across the entire internet. Step 8: Based on the user's basic information, including but not limited to geographical location, usage habits, education level, age, and spending power, push suitable search results to the user. Step 9: Record user habits and behavioral data as elements for subsequent service optimization. Step 10: Based on the data obtained in Step 9, pass it as an extended parameter to the search command when the user executes Step 6 again to participate in the next round of query. Effect illustration (see) Figure 2 ). Attached Figure Description

[0010] Figure 1 Business process diagram Figure 2 This is a diagram illustrating the business results.

Claims

1. A smart product recommendation method, characterized in that, The method includes the following steps: (1.1) Obtaining User Image: The platform obtains user appearance features and image information through hardware image acquisition devices. Alternatively, users can actively take photos using computer webcams, mobile phone webcams, or other photo-taking devices, upload images, and scan images to obtain user appearance characteristics and image information. (1.2) Select target scene: Users can select the scene preset by the platform or enter custom scene information, and add descriptions of scene features by clicking on tags or entering text. (1.3) Generate user styling suggestions: Use the information obtained in (1.1) and (1.2) as input parameters for the large-scale artificial intelligence model to automatically generate user styling suggestions. The suggestions include, but are not limited to, information on lifestyle products and services such as hairstyle, clothing, and accessories. (1.4) Based on (1.3) the platform recommends different levels of products and services to users according to the consumption amount, or pushes product information that meets the recommendation information characteristics of (1.3) to users through the Internet. (1.5) The platform continuously revises user preference settings based on user browsing, clicking, ordering, commenting and other behaviors, and continuously optimizes the user's search experience.

2. The intelligent product recommendation method according to claim 1, characterized in that, The user image collection in step (1.1) also includes user appearance features and image data obtained by any other image collection methods that are not prohibited by law and are authorized by the user.

3. The intelligent product recommendation method according to claim 1, characterized in that, The matching algorithm in step (1.4) also considers the time factor, that is, it dynamically adjusts the recommendation strategy according to the user's consumption habits in different time periods to achieve more accurate time-based recommendations.

4. The intelligent product recommendation method according to claim 1, characterized in that, The method also includes step (1.6) privacy protection mechanism, which means that encryption technology, anonymization and other means are used in the process of collecting, processing and recommending user behavior data to ensure the security of users' personal information and comply with relevant laws and regulations.

5. The intelligent product recommendation method according to any one of claims 1 to 4, characterized in that, This method can be applied to various network environments such as e-commerce platforms, social media platforms, and content delivery networks, providing personalized product recommendation services for users in different fields.

Images