A Product Recommendation Method and System Based on User Profiles

Abstract

The present application relates to the technical field of commodity recommendation, and in particular to a commodity recommendation method and system based on user portrait. The method includes: obtaining the browsing information of the user, analyzing the browsing information, extracting the user characteristic information; analyzing the user characteristic information, determining the user portrait information; determining the push content information according to the user portrait information; outputting the push content information for the user to browse and use. The present application analyzes the browsing information of the user, extracts the user characteristic information, determines the user portrait information on the basis of the user characteristic information, and determines the push content information according to the user portrait information, so as to output the push content information that is highly consistent with the user portrait for the user to browse and use, thereby improving the accuracy of the user portrait, making the push content information consistent with the actual needs of the user, improving the user experience, meeting the personalized needs of the user, and thus improving the user stickiness.

Description

Technical Field

[0001] The present application relates to the technical field of product recommendation, and in particular to a product recommendation method and system based on user portrait. Background Art

[0002] With the rapid development of the Internet and big data technology, the method of profiling users through their browsing and operation records on the Internet and pushing relevant information to users based on the user profiles has been widely used in various Internet platforms, playing a significant role in meeting users' personalized needs and improving user stickiness.

[0003] However, the existing information push method based on user portraits cannot accurately profile users during the process of pushing agricultural product information, resulting in the agricultural product push information not being able to meet the actual needs of users, which has a negative impact on user experience. Summary of the invention

[0004] The present application provides a product recommendation method and system based on user portrait to solve the above technical problems.

[0005] In a first aspect, the present application provides a method for recommending products based on user portraits, the method comprising:

[0006] Obtaining user browsing information, analyzing the browsing information, and extracting user feature information;

[0007] Analyze the user feature information to determine user portrait information;

[0008] Determine push content information according to the user portrait information;

[0009] The pushed content information is output for the user to browse and use.

[0010] During the process of users browsing agricultural product videos, this application obtains users' browsing information, analyzes the browsing information to extract user feature information, determines user portrait information based on the user feature information, and determines push content information based on the user portrait information, so as to output push content information that is highly consistent with the user portrait for users to browse and use, thereby improving the accuracy of user portraits, making the push content information consistent with users' actual needs, improving user experience, meeting users' personalized needs, and thereby improving user stickiness.

[0011] Optionally, the browsing information includes but is not limited to information about a number of agricultural product videos and browsing behavior information, and the analyzing the browsing information and extracting user characteristic information includes:

[0012] Analyze the video information of the agricultural products to determine video segment information;

[0013] Based on the video segment information and the browsing behavior information, determining skip information and stay information;

[0014] The skip information and the stay information are analyzed to determine preference information and career-oriented information, and the preference information and the career-oriented information are used as user feature information.

[0015] Through the above technical scheme, a number of agricultural product video information is analyzed to determine the video segmentation information, and then on the basis of the video segmentation information, the skip information and the stay information are determined according to the browsing behavior information, and the skip information and the stay information are analyzed to determine the preference information and the career orientation information. The preference information and the career orientation information are used as user characteristic information, so that the user portrait subsequently derived according to the user characteristic information is highly consistent with the user preference information, and the user portrait includes the user's career information, which further makes the user portrait more accurate.

[0016] Optionally, analyzing the agricultural product video information to determine video segmentation information includes:

[0017] Analyze the agricultural product video information to determine a number of scene change points;

[0018] Extracting a number of video segments according to the number of scene change points;

[0019] Analyze the plurality of video segments to determine the content type corresponding to each video segment;

[0020] Determine the total duration of each content type according to the plurality of agricultural product video information, the plurality of video segments and the content type;

[0021] The plurality of video segments, the content type and the total duration are determined as video segmentation information.

[0022] Through the above technical scheme, a number of agricultural product video information is analyzed, a number of scene change points are determined, and a number of video segments are extracted according to a number of scene change bags. Through the analysis of the several video segments, the content type corresponding to each video segment is determined. According to the several agricultural product video information, the several video segments and the content types, the total duration of each content type is determined. The several video segments, content types and total duration are determined as video segmentation information, which provides an important data basis for the subsequent determination of user stay information and skip information, making the determination of user characteristic information more accurate.

[0023] Optionally, analyzing the plurality of agricultural product video information to determine a plurality of scene change points includes:

[0024] Extracting frames from the video information of the agricultural products to determine a plurality of frames of images;

[0025] Analyze the plurality of frames of images and extract features of each frame of image;

[0026] According to the features of each frame of image, the preset feature difference value function and the preset difference threshold, several scene change points are determined by the following formula:

[0027] ;

[0028] in, is a set of several scene change points, For the Frame image features, For the Frame image features, is the preset feature difference value function, is the preset difference threshold, is the total number of frame images.

[0029] Through the above technical scheme, frame extraction is performed on a number of agricultural product video information, a number of frame images are determined, and the features of each frame image are extracted. The degree of change of the features of different frame images is measured by a mathematical formula to accurately determine a number of scene change points, so that the boundaries of a number of video segments obtained based on the number of scene change points are accurate.

[0030] Optionally, determining the total duration of each content type according to the plurality of agricultural product video information, the plurality of video segments and the content type includes:

[0031] Determining the number of agricultural product videos according to the agricultural product video information;

[0032] Determine the number of video segments and the duration of each video segment according to the plurality of video segments;

[0033] According to the number of agricultural product videos, the number of video segments, the duration of each video segment and the content type, the total duration of each content type is determined by referring to the following formula:

[0034] ;

[0035] in, For the content type The total duration of is the total number of video information of the agricultural products, is the total number of the video segments, For the Agricultural Products Video The duration of the video clip, For the Agricultural Products Video The content type of the video clip.

[0036] Through the above technical solution, a mathematical formula is designed to batch process the duration of each video clip in each agricultural product video based on the number of agricultural product videos, the number of video segments, the duration of each video segment and the content type, so as to accurately determine the total duration of each content type, providing accurate mathematical data basis for the subsequent determination of user stay information and skip information.

[0037] Optionally, the browsing behavior information includes but is not limited to a stay operation and a skip operation, and the determining the skip information and the stay information based on the video segment information and the browsing behavior information includes:

[0038] According to the stop operation, determining a number of stop periods;

[0039] According to the skip operation, determining a number of skip time periods;

[0040] Matching the plurality of stay periods and the plurality of skip periods with the plurality of video segments respectively, to determine a plurality of stay segments and a plurality of skip segments;

[0041] According to the plurality of stop paragraphs and the plurality of skip paragraphs, determining a stop paragraph content type corresponding to each stop paragraph and a skip paragraph content type corresponding to each skip paragraph;

[0042] According to the plurality of stay periods and the plurality of skip periods, respectively determine the total duration of each stay section content type and the total duration of each skip section content type;

[0043] Determine the duration ratio of each lingering paragraph content type and the duration ratio of each skipped paragraph content type according to the total duration of each lingering paragraph content type, the total duration of each skipped paragraph content type and the total duration;

[0044] Determine the duration ratio of all the content types of the stay sections as the stay information;

[0045] The skip information is determined as the percentage of the duration of all skipped paragraph content types.

[0046] Through the above technical scheme, a number of stay periods and a number of skip periods are determined respectively through the user's stay operations and skip operations, and the number of stay periods and the number of skip periods are matched with a number of video segments respectively to determine a number of stay segments and a number of skip segments, and then the paragraph content type corresponding to each stay segment and skip segment is determined. On the basis of the above, the total duration of each stay segment content type and the total duration of each skipped segment content type are determined, and then the duration ratio of each stay segment content type and the duration ratio of each skipped segment content type are obtained. The duration of each stay segment content type is used as stay information, and the duration ratio of each skipped segment content type is used as skip information, so as to simultaneously reflect a number of video clips and corresponding content types that the user focuses on and is not interested in, and further improve the accuracy of subsequent user portraits.

[0047] Optionally, the analyzing the skip information and the stay information to determine the preference information and the career orientation information includes:

[0048] Determine the preference information according to the stay segment content type having the highest proportion of duration among all stay segment content types;

[0049] Based on a preset occupational characteristic information database, a number of pre-selected occupational information are determined according to the skipped paragraph content type with the highest proportion of duration among all skipped paragraph content types;

[0050] Based on the proportion of the duration of the content types of all the stop segments, the plurality of pre-selected occupation information is analyzed to determine the occupation orientation information.

[0051] Through the above technical scheme, the user's preference information is determined according to the type of dwell paragraph content with the highest proportion in the duration of all dwell paragraph content types, so as to reflect the type of video clip content that the user pays close attention to. At the same time, the user's occupational information is screened according to the type of skipped paragraph content with the highest proportion in the duration of all skipped paragraph content types, and several pre-selected occupations are determined, so as to narrow the scope of analysis of the user's occupation and improve the speed and accuracy of obtaining occupation-oriented information. Then, according to the duration proportion of all dwell paragraph content types, several pre-selected occupations are further screened to determine accurate occupation-oriented information, so as to further improve the accuracy of user portraits.

[0052] Optionally, analyzing the user characteristic information to determine the user portrait information includes:

[0053] Analyze the preference information and the career orientation information to determine structured information indexed by the user's unique identifier;

[0054] The structured information is used as the user portrait information.

[0055] Through the above technical solution, after obtaining the user preference information and career orientation information, the user preference information and career orientation information are converted into structured information indexed by the user's unique identifier according to the structured information format, and the structured information is used as user portrait information, so that the user portrait information can be easily retrieved and analyzed, thereby improving the processing and calculation speed of the user portrait information, reducing the space occupied by the user portrait information, improving the response speed, and enhancing the user experience.

[0056] Optionally, determining the push content information according to the user portrait information includes:

[0057] Determining a number of pre-selected video clips according to the occupation-oriented information in the user portrait information;

[0058] Sorting the plurality of pre-selected video clips according to the preference information in the user portrait information to determine a sorted video clip list;

[0059] The sorted video clip list is used as the pushed content information.

[0060] Through the above technical solution, a number of pre-selected video clips are determined according to the career orientation information, and the pre-selected video clips are sorted according to the preference information to obtain a sorted video clip list, and the sorted video clip list is used as push content information, so that users can quickly obtain only the video clip information that meets their needs from the push content information, thereby further enhancing the user experience.

[0061] In a second aspect, the present application provides a product recommendation system based on user portraits, the system comprising:

[0062] A feature extraction module is used to obtain the user's browsing information, analyze the browsing information, and extract user feature information;

[0063] A portrait module, used to analyze the user characteristic information and determine the user portrait information;

[0064] A push content determination module, used to determine push content information according to the user portrait information;

[0065] The output module is used to output the pushed content information for the user to browse and use.

[0066] Optionally, the feature extraction module is specifically used to:

[0067] Analyze the video information of the agricultural products to determine video segment information;

[0068] Based on the video segment information and the browsing behavior information, determining skip information and stay information;

[0069] The skip information and the stay information are analyzed to determine preference information and career-oriented information, and the preference information and the career-oriented information are used as user feature information.

[0070] Optionally, the feature extraction module is specifically used to:

[0071] Analyze the agricultural product video information to determine a number of scene change points;

[0072] Extracting a number of video segments according to the number of scene change points;

[0073] Analyze the plurality of video segments to determine the content type corresponding to each video segment;

[0074] Determine the total duration of each content type according to the plurality of agricultural product video information, the plurality of video segments and the content type;

[0075] The plurality of video segments, the content type and the total duration are determined as video segmentation information.

[0076] Optionally, the feature extraction module is specifically used to:

[0077] Extracting frames from the video information of the agricultural products to determine a plurality of frames of images;

[0078] Analyze the plurality of frames of images and extract features of each frame of image;

[0079] According to the features of each frame of image, the preset feature difference value function and the preset difference threshold, several scene change points are determined by the following formula:

[0080] ;

[0081] in, is a set of several scene change points, For the Frame image features, For the Frame image features, is the preset feature difference value function, is the preset difference threshold, is the total number of frame images.

[0082] Optionally, the feature extraction module is specifically used to:

[0083] Determining the number of agricultural product videos according to the agricultural product video information;

[0084] Determine the number of video segments and the duration of each video segment according to the plurality of video segments;

[0085] According to the number of agricultural product videos, the number of video segments, the duration of each video segment and the content type, the total duration of each content type is determined by referring to the following formula:

[0086] ;

[0087] in, For the content type The total duration of is the total number of video information of the agricultural products, is the total number of the video segments, For the Agricultural Products Video The duration of the video clip, For the Agricultural Products Video The content type of the video clip.

[0088] Optionally, the feature extraction module is specifically used to:

[0089] According to the stop operation, determining a number of stop periods;

[0090] According to the skip operation, determining a number of skip time periods;

[0091] Matching the plurality of stay periods and the plurality of skip periods with the plurality of video segments respectively, to determine a plurality of stay segments and a plurality of skip segments;

[0092] According to the plurality of stop paragraphs and the plurality of skip paragraphs, determining a stop paragraph content type corresponding to each stop paragraph and a skip paragraph content type corresponding to each skip paragraph;

[0093] According to the plurality of stay periods and the plurality of skip periods, respectively determine the total duration of each stay section content type and the total duration of each skip section content type;

[0094] Determine the duration ratio of each lingering paragraph content type and the duration ratio of each skipped paragraph content type according to the total duration of each lingering paragraph content type, the total duration of each skipped paragraph content type and the total duration;

[0095] Determine the duration ratio of all the content types of the stay sections as the stay information;

[0096] The skip information is determined as the percentage of the duration of all skipped paragraph content types.

[0097] Optionally, the feature extraction module is specifically used to:

[0098] Determine the preference information according to the stay segment content type having the highest proportion of duration among all stay segment content types;

[0099] Based on a preset occupational characteristic information database, a number of pre-selected occupational information are determined according to the skipped paragraph content type with the highest proportion of duration among all skipped paragraph content types;

[0100] Based on the proportion of the duration of the content types of all the stop segments, the plurality of pre-selected occupation information is analyzed to determine the occupation orientation information.

[0101] Optionally, the portrait module is specifically used for:

[0102] Analyze the preference information and the career orientation information to determine structured information indexed by the user's unique identifier;

[0103] The structured information is used as the user portrait information.

[0104] Optionally, the push content determination module is specifically used to:

[0105] Determining a number of pre-selected video clips according to the occupation-oriented information in the user portrait information;

[0106] Sorting the plurality of pre-selected video clips according to the preference information in the user portrait information to determine a sorted video clip list;

[0107] The sorted video clip list is used as the pushed content information. BRIEF DESCRIPTION OF THE DRAWINGS

[0108] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0109] Figure 1 A schematic diagram of an application scenario provided for an embodiment of the present application;

[0110] Figure 2 A flowchart of a method for recommending products based on user portraits provided in one embodiment of the present application;

[0111] Figure 3 A schematic diagram of the structure of a product recommendation system based on user portraits provided in one embodiment of the present application. DETAILED DESCRIPTION

[0112] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0113] In addition, the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article, unless otherwise specified, generally means that the associated objects before and after are in an "or" relationship.

[0114] The embodiments of the present application are further described in detail below in conjunction with the drawings in the specification.

[0115] The existing information push method based on user portraits cannot accurately profile users during the process of pushing agricultural product information, resulting in the agricultural product push information not being able to meet the actual needs of users, which has a negative impact on user experience.

[0116] Based on this, the present application provides a commodity recommendation method and system based on user portrait. In the process of users browsing agricultural product videos, the user's browsing information is obtained, and the browsing information is analyzed to extract user feature information. Based on the user feature information, the user portrait information is determined, and the push content information is determined according to the user portrait information, so as to output push content information that is highly consistent with the user portrait for users to browse and use, thereby improving the accuracy of user portraits, making the push content information consistent with the actual needs of users, improving the user experience, meeting the personalized needs of users, and thus improving user stickiness.

[0117] Figure 1 A schematic diagram of an application scenario provided for the present application, in the process of a user browsing agricultural product video information, the method provided by the present application is applied to provide the user with corresponding push content information based on the user's browsing information.

[0118] Specifically, the method provided in the present application is applied to any server, which obtains the user's browsing information, analyzes the browsing information to extract user feature information, determines the user portrait information based on the user feature information, and determines the push content information based on the user portrait information, so as to output push content information that is highly consistent with the user portrait for the user to browse and use, thereby improving the accuracy of the user portrait, making the push content information consistent with the user's actual needs, improving the user experience, meeting the user's personalized needs, and thereby improving user stickiness.

[0119] For specific implementation methods, please refer to the following embodiments.

[0120] Figure 2 This is a flowchart of a method for recommending products based on user portraits provided in an embodiment of the present application. The method of this embodiment can be applied to the server in the above scenario. Figure 2 As shown, the method includes:

[0121] S201. Obtain user browsing information, analyze the browsing information, and extract user feature information.

[0122] The user can be considered as the target user browsing the agricultural product video.

[0123] Browsing information can be considered as key information involved in the process of users browsing agricultural product videos, such as agricultural product video information and browsing behavior information.

[0124] User characteristic information can be considered as relevant information about users' demand for agricultural products when they browse agricultural product videos, such as preference information and career orientation information.

[0125] Specifically, in the process of users browsing agricultural product videos, user behavior tracking tools such as Mixpanel and Mouseflow are used to track and extract the agricultural product video information browsed by users, as well as the behaviors performed by users in the process of browsing these agricultural product video information, such as fast forwarding and staying, to form browsing information. The browsing information is analyzed through image analysis and mathematical analysis to extract user feature information that can reflect the user's demand for agricultural products, providing an important data basis for subsequent accurate profiling of users based on user feature information.

[0126] S202: Analyze user feature information and determine user portrait information.

[0127] User portrait information can be considered as data that describes and classifies users in detail based on their personal characteristics, behavioral data, preferences, and other information.

[0128] Specifically, after determining the user characteristic information, the user's preference information for a certain type of agricultural products is determined by analyzing the user characteristic information, and the user's key focus information on this type of agricultural products is obtained. Based on the above information, the user is profiled and the user portrait information is determined to improve the accuracy of subsequent push content information determined based on the user portrait information.

[0129] S203: Determine push content information based on user portrait information.

[0130] The pushed content information can be considered as a list of relevant agricultural product videos that corresponds to the user portrait information and meets the user's actual needs.

[0131] Specifically, based on the information reflecting the user's actual needs in the user portrait information, several related agricultural product video clips are retrieved through the existing retrieval algorithm, and the several agricultural product video clips are integrated into a video list. The video list is used as push content information, so that the push content information is highly consistent with the user's actual needs, thereby enabling the user to quickly understand the relevant agricultural product information they need.

[0132] S204: Output push content information for users to browse and use.

[0133] Specifically, a push window is set on the user's browsing page, and the push content information is displayed in the push window to prevent the push content information from interfering with the user's normal browsing of other information, thereby improving the user's experience.

[0134] Through the method provided in this embodiment, in the process of users browsing agricultural product videos, the user's browsing information is obtained and analyzed to extract user feature information. Based on the user feature information, the user portrait information is determined, and the push content information is determined based on the user portrait information, so as to output push content information that is highly consistent with the user portrait for the user to browse and use, thereby improving the accuracy of the user portrait, making the push content information consistent with the user's actual needs, improving the user experience, meeting the user's personalized needs, and thereby improving user stickiness.

[0135] In some embodiments, several agricultural product video information is analyzed to determine video segmentation information; based on the video segmentation information and according to browsing behavior information, skip information and stay information are determined; the skip information and stay information are analyzed to determine preference information and career orientation information, and the preference information and career orientation information are used as user feature information.

[0136] Browsing information includes but is not limited to certain agricultural product video information and browsing behavior information.

[0137] Agricultural product video information can be considered as video information related to the introduction of agricultural products.

[0138] Browsing behavior information can be considered as the behavior performed by users when browsing agricultural product videos, such as fast forwarding and staying.

[0139] The video segmentation information can be considered as the video clip information corresponding to different video stages in the agricultural product video, such as the background introduction stage, the planting and harvesting stage, the processing stage, etc.

[0140] The skip information may be considered as information related to a video segment that a user skips by fast-forwarding or dragging a video progress bar.

[0141] The dwell information may be considered as information related to the video clip that the user is watching when the user does not perform a skip operation.

[0142] The preference information may be considered as the corresponding type of agricultural product video clips preferred by the user.

[0143] Career-oriented information can be considered as the user's career information obtained after analyzing the skipped information and the stayed information.

[0144] Specifically, when users browse agricultural product videos, they usually skip some video clips and focus on watching the video clips they are interested in. By analyzing the skipped clips and stayed clips of users, the users' interests can be reflected. Since the types of video clips in agricultural product videos are different, and different types of video clips are irregularly and unorderly interspersed with different time periods of agricultural product videos, firstly, image analysis technology is used to analyze the information of several agricultural product videos, and the agricultural product videos are segmented according to the change information between each clip, such as transition information, and then the video segmentation information is determined. On the basis of the video segmentation information, according to the user's browsing behavior information, it is determined which video clips the user skipped and which video clips the user stayed in, and then the skip information and stay information are obtained. According to the skip information and stay information, the user's preference information can be determined.

[0145] Furthermore, the characteristics of the video clips that the user skips and the characteristics of the video clips that the user stays on can map the user's occupational characteristics. For example, if the user skips other information and focuses on the planting and harvesting clips of agricultural product videos, the user is likely to be a grower. If the user focuses on the processing clips of agricultural products, the user is likely to be a purchasing staff of the processing plant. By analyzing the skipped information and the stayed information, the user's occupational orientation information can be determined. Using preference information and occupational orientation information as user characteristic information can significantly improve the accuracy of subsequent user portraits.

[0146] Through the method provided in this embodiment, a number of agricultural product video information is analyzed to determine the video segmentation information, and then on the basis of the video segmentation information, the skip information and the stay information are determined according to the browsing behavior information, and the skip information and the stay information are analyzed to determine the preference information and the career orientation information. The preference information and the career orientation information are used as user characteristic information, so that the user portrait subsequently derived according to the user characteristic information is highly consistent with the user preference information, and the user portrait includes the user's career information, which further makes the user portrait more accurate.

[0147] In some embodiments, a number of agricultural product video information is analyzed to determine a number of scene change points; based on the number of scene change points, a number of video segments are extracted; the number of video segments are analyzed to determine the content type corresponding to each video segment; based on the number of agricultural product video information, the number of video segments and the content type, the total duration of each content type is determined; the number of video segments, the content type and the total duration are determined as video segmentation information.

[0148] The scene change points can be considered as the switching time points of different types of video clips in agricultural product videos.

[0149] Video segments can be considered as different video segments obtained by dividing the agricultural product video according to scene change points.

[0150] The content type can be considered as the type corresponding to the content of the video segment, such as background introduction type, planting and harvesting type, processing type, market analysis type, etc.

[0151] The total duration of each content type can be considered as the total duration of the video clips corresponding to each content type in the agricultural product video.

[0152] Specifically, in the process of determining the video segmentation information, the agricultural product video is segmented according to the scene change nodes or editing nodes in the agricultural product video through video segmentation tools, such as PySceneDetect, to determine a number of video segments, and then the images in the video segments are analyzed through computer vision processing tools, such as OpenCV (Open Source Computer Vision Library) to determine the content type corresponding to each video segment, and on the basis of a number of agricultural product video information, a number of video segments and content types, the total duration of each content type is determined through mathematical analysis, and the number of video segments, content types and total duration are determined as video segmentation information, which provides an important data basis for the subsequent determination of user stay information and skip information.

[0153] Through the method provided in this embodiment, a number of agricultural product video information is analyzed, a number of scene change points are determined, and a number of video segments are extracted according to the number of scene change bags. By analyzing the number of video segments, the content type corresponding to each video segment is determined. According to the number of agricultural product video information, the number of video segments and the content type, the total duration of each content type is determined. The number of video segments, content types and total duration are determined as video segmentation information, which provides an important data basis for the subsequent determination of user stay information and skip information, making the determination of user feature information more accurate.

[0154] In some embodiments, frame extraction is performed on a number of agricultural product video information to determine a number of image frames; the number of image frames are analyzed to extract features of each image frame; and based on the features of each image frame, a preset feature difference value function and a preset difference threshold, a number of scene change points are determined by formula (1):

[0155] (1);

[0156] in, is a set of several scene change points, For the Frame image features, For the Frame image features, is the preset feature difference value function, is the preset difference threshold, is the total number of frame images.

[0157] The frame image can be considered as the image corresponding to each frame in each agricultural product video information.

[0158] The frame image feature can be considered as information used to express the frame image feature, and the frame image feature can be expressed using the corresponding frame image vector feature information.

[0159] The preset feature difference function can be considered as a preset function for calculating the degree of change between features of different frame images. The preset feature difference function can adopt an existing vector feature change degree measurement algorithm, such as a Euclidean distance algorithm and a similarity measurement algorithm.

[0160] The preset difference threshold can be considered as the maximum value of the degree of change between frame image features in the same scene. When the degree of change between frame image features exceeds the preset difference threshold, it can be considered that the scene has changed. The preset difference threshold can be obtained by analyzing historical data.

[0161] Specifically, in the process of determining several scene change points, firstly, a computer vision processing tool, such as OpenCV, is used to extract frames of agricultural product video information to determine several frames of images. Then, a convolutional neural network, such as VGG (Visual Geometry Group) and ResNet (Residual Network), is used to extract feature vectors of several frames of images to determine the features of each frame of images. The degree of change between different frames of images is measured by formula (1) to determine several scene change points.

[0162] Through the method provided in this embodiment, frame extraction is performed on a number of agricultural product video information, a number of frame images are determined, and features of each frame image are extracted. The degree of change of features of different frame images is measured by a mathematical formula to accurately determine a number of scene change points, so that the boundaries of a number of video segments obtained according to the number of scene change points are accurate.

[0163] In some embodiments, the number of agricultural product videos is determined based on a number of agricultural product video information; the number of video segments and the duration of each video segment are determined based on a number of video segments; and the total duration of each content type is determined based on the number of agricultural product videos, the number of video segments, the duration of each video segment and the content type, with reference to formula (2):

[0164] (2);

[0165] in, For content type The total duration of is the total number of video information of several agricultural products, is the total number of video segments, For the Agricultural Products Video The duration of the video clip, For the Agricultural Products Video The content type of the video clip.

[0166] Specifically, the duration of each video segment in each agricultural product video is batch processed according to the number of agricultural product videos, the number of video segments, the duration of each video segment, and the content type through formula (2), where is the indicator function, when hour, The value of is 1, otherwise, The value of is 0 to determine the total duration of each content type.

[0167] Through the method provided in this embodiment, a mathematical formula is designed to batch process the duration of each video segment in each agricultural product video based on the number of agricultural product videos, the number of video segments, the duration of each video segment and the content type, so as to accurately determine the total duration of each content type, thereby providing accurate mathematical data basis for the subsequent determination of user stay information and skip information.

[0168] In some embodiments, based on the stop operation, a number of stop periods are determined; based on the skip operation, a number of skip periods are determined; the number of stop periods and the number of skip periods are matched with a number of video segments respectively to determine a number of stop segments and a number of skip segments; based on the number of stop segments and the number of skip segments, the stop segment content type corresponding to each stop segment and the skip segment content type corresponding to each skip segment are determined; based on the number of stop periods and the number of skip periods, the total duration of each stop segment content type and the total duration of each skip segment content type are determined respectively; based on the total duration of each stop segment content type, the total duration of each skipped segment content type and the total duration, the duration ratio of each stop segment content type and the duration ratio of each skipped segment content type are determined; the duration ratio of all stop segment content types are determined as the stop information; the duration ratio of all skipped segment content types are determined as the stop information.

[0169] Browsing behavior information includes but is not limited to stay operations and skip operations.

[0170] The dwell operation can be considered as the operation of users lingering on a certain video clip while browsing agricultural product videos.

[0171] The skip operation can be considered as the user's operation of skipping some video clips when browsing agricultural product videos.

[0172] The dwell segment can be considered as the video segment corresponding to the user's dwell operation.

[0173] The skipped section can be considered as the video segment corresponding to the user's skip operation.

[0174] The total duration of a content type of a dwell segment can be considered as the total duration of a user's stay in each type of dwell segment.

[0175] The total duration of the skipped paragraph content type can be considered as the total skipping duration of the user in each type of skipped paragraph.

[0176] The proportion of the dwell segment content type duration can be considered as the proportion of the total duration of each dwell segment content type in the total duration of each corresponding content type.

[0177] The skipped paragraph content type duration ratio can be considered as the ratio of the total duration of each skipped paragraph content type to the total duration of each corresponding content type.

[0178] Specifically, through operation log monitoring technology, according to the user's stay operations and skip operations, the timestamp information corresponding to the user's stay operations and skip operations is extracted to determine the corresponding several stay periods and skip periods of the user in the process of browsing agricultural products videos, and then according to the several stay periods and skip periods, the video segments corresponding to their time periods are matched in several video segments to obtain several stay segments and several skip segments, and then determine the corresponding content type of each stay segment and skip segment.

[0179] Furthermore, based on a number of dwell periods, a number of skip periods and corresponding content types, the total duration of each dwell paragraph content type and the total duration of each skip paragraph content type are determined respectively, and based on the corresponding total duration of each content type, the duration ratio of each dwell paragraph content type and the duration ratio of each skip paragraph content type are determined. By determining the duration ratio of all dwell paragraph content types as the dwell information, several video clips and corresponding content types that the user focuses on are reflected. At the same time, by determining the duration ratio of all skip paragraph content types as the skip information, several video clips and corresponding content types that the user is not interested in are reflected.

[0180] The method provided in this embodiment determines a number of stay periods and a number of skip periods respectively through the user's stay operations and skip operations, and matches the number of stay periods and the number of skip periods with a number of video segments respectively to determine a number of stay segments and a number of skip segments, and then determines the paragraph content type corresponding to each stay segment and skip segment, and on the above basis, determines the total duration of each stay segment content type and the total duration of each skip segment content type, and then obtains the duration ratio of each stay segment content type and the duration ratio of each skip segment content type, and uses the duration of each stay segment content type as stay information, and uses the duration ratio of each skip segment content type as skip information, so as to simultaneously reflect the number of video clips and corresponding content types that the user focuses on and is not interested in, and further improves the accuracy of subsequent user portraits.

[0181] In some embodiments, preference information is determined based on the type of content of the stop paragraphs with the highest proportion in duration among all types of content of the stop paragraphs; based on a preset occupational characteristic information database, a number of pre-selected occupational information are determined based on the type of content of the skip paragraphs with the highest proportion in duration among all types of content of the skip paragraphs; based on the duration proportion of all types of content of the stop paragraphs, a number of pre-selected occupational information are analyzed to determine occupational orientation information.

[0182] The preset occupational feature information database can be considered as an information database for storing the relationship between paragraph content types and user occupations.

[0183] The pre-selected occupational information can be considered as information on several occupations that the user may engage in, obtained by sorting out a series of occupational information in a preset occupational feature information database based on the type of content that is skipped by the user with the highest proportion of the content time.

[0184] Specifically, users' preference information for agricultural product videos can be reflected from the content types of the segments where users stay. The content types of the segments with the highest proportion of the user's stay duration are highly representative, reflecting which content types the user pays close attention to in agricultural product videos. Using this content type as preference information can improve the accuracy of user portraits.

[0185] At the same time, the type of content that is skipped with the highest proportion in duration among all content types that are skipped by users can reflect the type of content that users are least interested in, thereby ruling out the possibility that the user is engaged in certain occupations. For example, if the length of time that users skip over planting and harvesting video segments within agricultural product videos is the highest, it can indicate that the user is not a grower, thereby narrowing the scope of analysis of the user's occupation and improving the speed and accuracy of obtaining occupation-oriented information.

[0186] Furthermore, after obtaining several pre-selected occupational information, the pre-selected occupational information is further screened through the proportion of the time that users stay on all the content types in the paragraphs. Users of different occupations have different focuses when browsing agricultural product videos, which makes the proportion of the time that users of different occupations stay on the content types in the paragraphs different. By analyzing historical data or large-scale questionnaires, the proportion of the time that users of different occupations stay on all the content types in the paragraphs when browsing agricultural product videos can be collected. This can be used as a basis for judgment. According to the proportion of the time that the current user stays on all the content types in the paragraphs, the user's occupational orientation information can be accurately obtained.

[0187] Through the method provided in this embodiment, the user's preference information is determined based on the dwell paragraph content type with the highest proportion in the duration of all dwell paragraph content types, so as to reflect the video clip content type that the user is highly concerned about. At the same time, the user's occupational information is screened according to the skipped paragraph content type with the highest proportion in the duration of all skipped paragraph content types, and a number of pre-selected occupations are determined, so as to narrow the scope of analysis of the user's occupation and improve the speed and accuracy of obtaining occupation-oriented information. Then, according to the duration proportion of all dwell paragraph content types, a number of pre-selected occupations are further screened to determine accurate occupation-oriented information, so as to further improve the accuracy of user portraits.

[0188] In some embodiments, preference information and career-oriented information are analyzed to determine structured information indexed by the user's unique identifier; and the structured information is used as user portrait information.

[0189] The user's unique identifier can be considered as identification information representing the user's identity. The user's unique identifier can be the unique identifier of the device hardware used by the user, or it can be assigned a unique identifier to the user through a unique identification algorithm, such as UUID (Universally Unique Identifier) ​​and Snowflake (snowflake algorithm).

[0190] Structured information can be considered as information used to organize, classify and associate user unique identification, preference information and career-oriented information according to established rules and formats. Structured information can be easily processed and understood automatically by computer programs. The structured information format can use JSON (JavaScript Object Notation).

[0191] Specifically, after obtaining the user preference information and career orientation information, the user preference information and career orientation information are converted into structured information indexed by the user's unique identifier according to the structured information format, and the structured information is used as the user portrait information.

[0192] Through the method provided in this embodiment, after obtaining the user preference information and career orientation information, the user preference information and career orientation information are converted into structured information indexed by the user's unique identifier according to the structured information format, and the structured information is used as user portrait information, so that the user portrait information can be easily retrieved and analyzed, thereby improving the processing and calculation speed of the user portrait information, reducing the space occupied by the user portrait information, and improving the response speed, thereby enhancing the user experience.

[0193] In some embodiments, a number of pre-selected video clips are determined based on the career-oriented information in the user portrait information; the number of pre-selected video clips are sorted based on the preference information in the user portrait information to determine a sorted video clip list; and the sorted video clip list is used as push content information.

[0194] The pre-selected video clips may be considered to be a number of video clips related to the user's occupation obtained by retrieving information directed to the user's occupation.

[0195] The sorted video clip list can be considered as a list of video clips sorted by the degree of conformity with the user's preference information after sorting the pre-selected video clips according to the user's preference information.

[0196] Specifically, the existing retrieval algorithm is used to retrieve relevant video clips of the user's occupation according to the user's occupation information, and several pre-selected video clips are determined. However, since the relevant video clips of the occupation cover a wide range, the user cannot quickly locate the required video information. Therefore, the several pre-selected video clips are sorted according to the user's preference information, and the order of the video clips with a high match with the user's preference information is adjusted forward to obtain a sorted video clip list, and the sorted video clip list is used as the push content information.

[0197] In some other embodiments, when there are few existing video clips related to the user's occupation, several agricultural product videos are analyzed through a series of methods in the aforementioned embodiments, and the starting timestamps of the video clips related to the user's occupation are extracted. When the user clicks on the agricultural product video, the video automatically jumps to the corresponding progress of the agricultural product video according to the starting timestamp, so that the user can quickly obtain the required video clip information.

[0198] Through the method provided in this embodiment, a number of pre-selected video clips are determined according to the career orientation information, and the pre-selected video clips are sorted according to the preference information to obtain a sorted video clip list, and the sorted video clip list is used as push content information, so that users can quickly obtain only the video clip information that meets their needs from the push content information, thereby further enhancing the user experience.

[0199] Figure 3 A schematic diagram of a product recommendation system based on user portraits provided in an embodiment of the present application is shown in FIG. Figure 3 As shown, the product recommendation system 300 based on user portrait in this embodiment includes: a feature extraction module 301 , a portrait module 302 , a push content determination module 303 and an output module 304 .

[0200] The feature extraction module 301 is used to obtain the browsing information of the user, analyze the browsing information, and extract the user feature information;

[0201] A portrait module 302 is used to analyze the user characteristic information and determine user portrait information;

[0202] A push content determination module 303 is used to determine push content information according to the user portrait information;

[0203] The output module 304 is used to output the pushed content information for the user to browse and use.

[0204] Optionally, the feature extraction module 301 is specifically used for:

[0205] Analyze the video information of the agricultural products to determine video segment information;

[0206] Based on the video segment information and the browsing behavior information, determining skip information and stay information;

[0207] The skip information and the stay information are analyzed to determine preference information and career-oriented information, and the preference information and the career-oriented information are used as user feature information.

[0208] Optionally, the feature extraction module 301 is specifically used for:

[0209] Analyze the agricultural product video information to determine a number of scene change points;

[0210] Extracting a number of video segments according to the number of scene change points;

[0211] Analyze the plurality of video segments to determine the content type corresponding to each video segment;

[0212] Determine the total duration of each content type according to the plurality of agricultural product video information, the plurality of video segments and the content type;

[0213] The plurality of video segments, the content type and the total duration are determined as video segmentation information.

[0214] Optionally, the feature extraction module 301 is specifically used for:

[0215] Extracting frames from the video information of the agricultural products to determine a plurality of frames of images;

[0216] Analyze the plurality of frames of images and extract features of each frame of image;

[0217] According to the features of each frame of image, the preset feature difference value function and the preset difference threshold, several scene change points are determined by the following formula:

[0218] ;

[0219] in, is a set of several scene change points, For the Frame image features, For the Frame image features, is the preset feature difference value function, is the preset difference threshold, is the total number of frame images.

[0220] Optionally, the feature extraction module 301 is specifically used for:

[0221] Determining the number of agricultural product videos according to the agricultural product video information;

[0222] Determine the number of video segments and the duration of each video segment according to the plurality of video segments;

[0223] According to the number of agricultural product videos, the number of video segments, the duration of each video segment and the content type, the total duration of each content type is determined by referring to the following formula:

[0224] ;

[0225] in, For the content type The total duration of is the total number of video information of the agricultural products, is the total number of the video segments, For the Agricultural Products Video The duration of the video clip, For the Agricultural Products Video The content type of the video clip.

[0226] Optionally, the feature extraction module 301 is specifically used for:

[0227] According to the stop operation, determining a number of stop periods;

[0228] According to the skip operation, determining a number of skip time periods;

[0229] Matching the plurality of stay periods and the plurality of skip periods with the plurality of video segments respectively, to determine a plurality of stay segments and a plurality of skip segments;

[0230] According to the plurality of stop paragraphs and the plurality of skip paragraphs, determining a stop paragraph content type corresponding to each stop paragraph and a skip paragraph content type corresponding to each skip paragraph;

[0231] According to the plurality of stay periods and the plurality of skip periods, respectively determine the total duration of each stay section content type and the total duration of each skip section content type;

[0232] Determine the duration ratio of each lingering paragraph content type and the duration ratio of each skipped paragraph content type according to the total duration of each lingering paragraph content type, the total duration of each skipped paragraph content type and the total duration;

[0233] Determine the duration ratio of all the content types of the stay sections as the stay information;

[0234] The skip information is determined as the percentage of the duration of all skipped paragraph content types.

[0235] Optionally, the feature extraction module 301 is specifically used for:

[0236] Determine the preference information according to the stay segment content type having the highest proportion of duration among all stay segment content types;

[0237] Based on a preset occupational characteristic information database, a number of pre-selected occupational information are determined according to the skipped paragraph content type with the highest proportion of duration among all skipped paragraph content types;

[0238] Based on the proportion of the duration of the content types of all the stop segments, the plurality of pre-selected occupation information is analyzed to determine the occupation orientation information.

[0239] Optionally, the portrait module 302 is specifically used for:

[0240] Analyze the preference information and the career orientation information to determine structured information indexed by the user's unique identifier;

[0241] The structured information is used as the user portrait information.

[0242] Optionally, the push content determination module 303 is specifically used to:

[0243] Determining a number of pre-selected video clips according to the occupation-oriented information in the user portrait information;

[0244] Sorting the plurality of pre-selected video clips according to the preference information in the user portrait information to determine a sorted video clip list;

[0245] The sorted video clip list is used as the pushed content information.

[0246] The system of this embodiment can be used to execute the method of any of the above embodiments. The implementation principles and technical effects are similar and will not be described in detail here.

Claims

1. A commodity recommendation method based on user portrait, characterized in that: include: Obtaining user browsing information, analyzing the browsing information, and extracting user feature information; Analyze the user feature information to determine user portrait information; Determine push content information according to the user portrait information; Outputting the pushed content information for the user to browse and use; The browsing information includes a number of agricultural product video information and browsing behavior information, and the analyzing the browsing information and extracting user characteristic information includes: Analyze the video information of the agricultural products to determine video segment information; Based on the video segment information and the browsing behavior information, determining skip information and stay information; Analyze the skip information and the stay information to determine preference information and career orientation information, and use the preference information and the career orientation information as user feature information; The step of analyzing the agricultural product video information and determining the video segmentation information includes: Analyze the agricultural product video information to determine a number of scene change points; Extracting a number of video segments according to the number of scene change points; Analyze the plurality of video segments to determine the content type corresponding to each video segment; Determine the total duration of each content type according to the plurality of agricultural product video information, the plurality of video segments and the content type; Determine the plurality of video segments, the content type, and the total duration as video segmentation information; The browsing behavior information includes a stay operation and a skip operation, and the determining the skip information and the stay information based on the video segment information and the browsing behavior information includes: According to the stop operation, determining a number of stop periods; According to the skip operation, determining a number of skip time periods; Matching the plurality of stay periods and the plurality of skip periods with the plurality of video segments respectively, to determine a plurality of stay segments and a plurality of skip segments; According to the plurality of stop paragraphs and the plurality of skip paragraphs, determining a stop paragraph content type corresponding to each stop paragraph and a skip paragraph content type corresponding to each skip paragraph; According to the plurality of stay periods and the plurality of skip periods, respectively determine the total duration of each stay section content type and the total duration of each skip section content type; Determine the duration ratio of each lingering paragraph content type and the duration ratio of each skipped paragraph content type according to the total duration of each lingering paragraph content type, the total duration of each skipped paragraph content type and the total duration; Determine the duration ratio of all the content types of the stay sections as the stay information; Determine the skip information as the percentage of the content type duration of all skipped paragraphs; The analyzing the skip information and the stay information to determine the preference information and the career orientation information includes: Determine the preference information according to the stay segment content type having the highest proportion of duration among all stay segment content types; Based on a preset occupational characteristic information database, a number of pre-selected occupational information are determined according to the skipped paragraph content type with the highest proportion of duration among all skipped paragraph content types; Based on the proportion of the duration of the content types of all the stop segments, the plurality of pre-selected occupation information is analyzed to determine the occupation orientation information.

2. The method according to claim 1, characterized in that The analyzing the plurality of agricultural product video information to determine a plurality of scene change points includes: Extracting frames from the video information of the agricultural products to determine a plurality of frames of images; Analyze the plurality of frames of images and extract features of each frame of image; According to the features of each frame of image, the preset feature difference value function and the preset difference threshold, several scene change points are determined by the following formula: ； Wherein, B is a set of several scene change points, is the image feature of the i-th frame, is the image feature of the i+1th frame, is the preset feature difference value function, is the preset difference threshold, and N is the total number of frame images.

3. The method according to claim 1, characterized in that The determining the total duration of each content type according to the plurality of agricultural product video information, the plurality of video segments and the content type includes: Determining the number of agricultural product videos according to the agricultural product video information; Determine the number of video segments and the duration of each video segment according to the plurality of video segments; According to the number of agricultural product videos, the number of video segments, the duration of each video segment and the content type, the total duration of each content type is determined by referring to the following formula: ； in, is the total duration of the content type C, N is the total number of the agricultural product video information, is the total number of the plurality of video segments, is the duration of the jth video clip of the i-th agricultural product video, is the content type of the jth video segment of the i-th agricultural product video.

4. The method according to claim 1, characterized in that: The analyzing the user characteristic information to determine the user portrait information includes: Analyze the preference information and the career orientation information to determine structured information indexed by the user's unique identifier; The structured information is used as the user portrait information.

5. The method according to claim 4, characterized in that The step of determining push content information according to the user portrait information includes: Determining a number of pre-selected video clips according to the occupation-oriented information in the user portrait information; Sorting the plurality of pre-selected video clips according to the preference information in the user portrait information to determine a sorted video clip list; The sorted video clip list is used as the pushed content information.

6. A product recommendation system based on user portrait, characterized in that: The method as claimed in any one of claims 1 to 5 comprises: A feature extraction module is used to obtain the user's browsing information, analyze the browsing information, and extract user feature information; A portrait module, used to analyze the user characteristic information and determine the user portrait information; A push content determination module, used to determine push content information according to the user portrait information; The output module is used to output the pushed content information for the user to browse and use.

Images