An advertisement recommendation effect evaluation system based on multi-dimensional user portraits

Abstract

The application discloses a kind of advertisement recommendation effect evaluation systems based on multidimensional user portrait, it is related to artificial intelligence technical field, the functional module of the system includes: user portrait quantification module, space-time correction module, sample weight correction module, joint evaluation score calculation module and normalization quantification output module.The application is measured by setting the user portrait abnormal risk measurement mechanism based on space-time correction, and is normalized modeling by joint sample weight, forms unified abnormal risk joint evaluation score, and the scheme effectively overcomes the deficiency of abnormal risk isolated evaluation in the prior art, improves the stability and credibility of advertisement recommendation effect evaluation.The technical means provides a solid data basis for risk identification and effect quantification of recommendation model, and promotes the development of advertisement recommendation system to more accurate and intelligent direction.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, in particular to an advertisement recommendation effect evaluation system based on multi-dimensional user portrait. BACKGROUND

[0002] With the popularity of the Internet and mobile terminals, advertisement recommendation systems play an important role in digital marketing. Accurate evaluation of advertisement recommendation effects is not only directly related to the rational allocation of marketing resources, but also affects user experience and platform revenue. Traditional advertisement recommendation effect evaluation relies on single-dimensional user behavior data or simple statistical indicators, which is difficult to fully reflect the changes in user interest and the actual impact of advertisements. In recent years, multi-dimensional user portrait technology has been gradually applied to advertisement recommendation, which attempts to achieve more accurate user interest portrayal and recommendation effect judgment by integrating user behavior data, geographic location, time characteristics and other multiple dimensions.

[0003] In the prior art, although some methods attempt to introduce user portrait spatiotemporal information for recommendation risk evaluation, there is a common problem of evaluating sample abnormal scores separately, lacking joint modeling of sample weights, resulting in insufficient robustness and accuracy of the overall evaluation. SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides an advertisement recommendation effect evaluation system based on multi-dimensional user portrait to solve the problems raised in the background art.

[0005] To achieve the above purpose, the present application provides the following technical solutions:

[0006] In a first aspect, the present application provides an advertisement recommendation effect evaluation system based on multi-dimensional user portrait, the functional modules of the system including: a user portrait quantification module, a spatiotemporal correction module, a sample weight correction module, a joint evaluation score calculation module and a normalized quantization output module.

[0007] The functional modules of the system include the following steps:

[0008] Based on the user portrait quantification module, the attributes of the collected user portrait are quantified to obtain a portrait vector.

[0009] The spatiotemporal correction module is used to correct the portrait vector obtained by the user portrait quantification module in space and time to obtain a spatiotemporal correction portrait vector.

[0010] The sample weight correction module is used to correct the spatiotemporal correction portrait vector output by the spatiotemporal correction module in sample weight to obtain a sample weight.

[0011] Based on the sample weight correction module output sample weight, using joint evaluation score calculation module for joint evaluation, get joint evaluation score;

[0012] The joint evaluation score calculation module output joint evaluation score input normalization quantization output module, normalization quantization, get quantization output score, as the basis for measuring the risk of advertising recommendation, the higher the score, the higher the risk of advertising recommendation.

[0013] Further optimize the technical solution, the input portrait in the user portrait quantization module contains: user basic information, user interest label, user behavior data and user geographic information, through One-Hot coding technology, WordEmbedding technology and normalization processing on the above portrait information quantization, form portrait vector .

[0014] Further optimize the technical solution, the space-time correction module first carries out space-time weight modeling, then carries out space-time deviation correction, and finally carries out corrected portrait reconstruction;

[0015] In the space-time weight modeling, the space-time weight is applied to each portrait component, and the space-time weight is determined by the space-time factor. The corresponding space-time weight modeling is a space-time function, and the formula is:

[0016] ;

[0017] Among them,

[0018] : space-time weight;

[0019] : collection time;

[0020] : collection position;

[0021] : Sigmoid activation function;

[0022] : sensitivity of space-time weight to time;

[0023] : sensitivity of space-time weight to position;

[0024] : bias, ensure the stability of weight distribution.

[0025] Further optimize the technical solution, the space-time correction module eliminates the deviation of portrait occurrence caused by space-time conditions when performing space-time deviation correction, and realizes it by correcting each component. The corresponding model formula is:

[0026] ;

[0027] in,

[0028] : Corrected Each portrait component;

[0029] : No. before correction Each portrait component;

[0030] Spatiotemporal weights are used to perform multiplicative corrections on each component.

[0031] To further optimize this technical solution, the spatiotemporal correction module uses the following reconstruction formula when reconstructing the corrected image:

[0032] ;

[0033] in,

[0034] : The new image vector formed after spatiotemporal correction.

[0035] To further optimize this technical solution, the sample weight correction module first performs sample weight modeling, then performs sample weight correction, and finally outputs the weights. The abnormal score is:

[0036] ;

[0037] in,

[0038] : No. Spatiotemporal corrected image vector for each sample;

[0039] The equilibrium vector obtained from the overall mean of the spatiotemporally corrected image. In the formula The total number of samples.

[0040] To further optimize this technical solution, the sample weight correction module assigns weights to each sample based on the outlier score when performing sample weight correction:

[0041] ;

[0042] in, Sample weights;

[0043] :sample Abnormal scores;

[0044] Weight decay factor.

[0045] To further optimize this technical solution, the sample weight correction module ultimately outputs the weight of each sample:

[0046] ;

[0047] And this forms the sample weights:

[0048] .

[0049] To further optimize this technical solution, the joint evaluation score calculation module first performs weight normalization, then models the joint evaluation score to obtain the joint evaluation score. .

[0050] In a second aspect, embodiments of the present invention provide a computer device, including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program instructions are executed by the processor, they implement the steps of an advertising recommendation effect evaluation system based on multi-dimensional user profiles as described in the first aspect of the present invention.

[0051] Thirdly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program instructions are executed by a processor, they implement the steps of an advertising recommendation effect evaluation system based on a multi-dimensional user profile as described in the first aspect of the present invention.

[0052] Compared with existing technologies, this invention provides an advertising recommendation effectiveness evaluation system based on multi-dimensional user profiles, which has the following beneficial effects:

[0053] This advertising recommendation effectiveness evaluation system, based on multi-dimensional user profiles, establishes a user profile anomaly risk measurement mechanism based on spatiotemporal correction and performs normalization modeling using sample weights to form a unified joint evaluation score for anomaly risk. This solution effectively overcomes the shortcomings of isolated anomaly risk evaluation in existing technologies, improving the stability and reliability of advertising recommendation effectiveness evaluation. This technical approach provides a solid data foundation for risk identification and effectiveness quantification in recommendation models, driving advertising recommendation systems towards greater accuracy and intelligence. Attached Figure Description

[0054] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0055] Figure 1This is a flowchart illustrating an advertising recommendation effectiveness evaluation system based on multi-dimensional user profiles proposed in this invention.

[0056] Figure 2 This is a schematic diagram of the spatiotemporal correction process for an advertising recommendation effect evaluation system based on multi-dimensional user profiles proposed in this invention.

[0057] Figure 3 This is a schematic diagram of the sample weight correction process for an advertising recommendation effect evaluation system based on multi-dimensional user profiles proposed in this invention.

[0058] Figure 4 This is a schematic diagram of the joint evaluation process of an advertising recommendation effect evaluation system based on multi-dimensional user profiles proposed in this invention.

[0059] Figure 5 This is a schematic diagram of the functional modules of an advertising recommendation effect evaluation system based on multi-dimensional user profiles proposed in this invention. Detailed Implementation

[0060] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0061] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0062] Secondly, the term "an embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places throughout this specification does not necessarily refer to the same embodiment, nor is it a single embodiment or an embodiment selectively excluded from other embodiments.

[0063] Example 1:

[0064] Reference Figures 1-4 This is the first embodiment of the present invention, which provides an advertising recommendation effectiveness evaluation system based on multi-dimensional user profiles, including the following steps:

[0065] S1. Use the collected user profiles to quantify attributes and obtain profile vectors;

[0066] The multi-dimensional user profile data input in step S1 mainly comes from user behavior collection devices on various terminals such as mobile terminals, PCs, and IoT devices, including but not limited to application clients, web browsers, positioning sensors, and social platform interfaces. After collecting multi-dimensional information such as user click behavior, browsing history, geographic location, timestamps, and interest tags through data acquisition technology, data cleaning and preprocessing techniques are used to denoise, convert formats, and extract features from the raw data. Finally, numerical encoding and vectorization techniques are used to transform the multi-dimensional discrete and continuous attributes into structured user profile vectors, providing standardized and computable profile input for subsequent steps.

[0067] The input profile for step S1 includes:

[0068] Basic user information, including age, gender, and other information tags;

[0069] User interest tags, including category tags such as sports, movies, and games;

[0070] User behavior data, including numerical information such as browsing time and number of clicks;

[0071] User geographic information, i.e., location information.

[0072] Step S1 uses One-Hot encoding technology to quantify the classification attributes such as "gender" and "region" in the above portrait information. Gender is "male" or "female", which are One-Hot encoded as "[1,0]" and "[0,1]" respectively. The region information is One-Hot encoded according to the administrative region.

[0073] Word Embedding technology is used to embed each interest tag. If a user has multiple tags, the final interest embedding vector is obtained by vectorizing each tag.

[0074] ;

[0075] For user behavior data, numerical information such as browsing time and number of clicks are normalized to form decimals with a field of [0,1] to eliminate differences in dimensions.

[0076] Finally, these quantified pieces of information are concatenated to form a profile vector. .

[0077] S2. Use the image vector to perform spatiotemporal correction to obtain the spatiotemporally corrected image vector;

[0078] Step S2 is based on the image vector obtained in step S1. Spatiotemporal correction operations are performed, including spatiotemporal weight modeling, spatiotemporal bias correction, and reconstruction of the corrected image.

[0079] Step S2 first performs spatiotemporal weight modeling, applying spatiotemporal weights to each image component. These weights are determined by spatiotemporal factors, and the corresponding spatiotemporal weights are modeled as a spatiotemporal function, with the following formula:

[0080] ;

[0081] in,

[0082] Spatiotemporal weights;

[0083] : Data collection time;

[0084] : Data collection location;

[0085] : Sigmoid activation function;

[0086] The sensitivity of spatiotemporal weights to time is determined by training historical samples using existing data fitting techniques. This process includes regression analysis of the relationship between spatiotemporal conditions at different times and recommendation risk, to determine which value best measures the effect of time on spatiotemporal weights.

[0087] The sensitivity of spatiotemporal weights to location is determined by regression analysis of risk samples from different geographical locations to identify the optimal fitting parameters when location affects spatiotemporal weights.

[0088] The bias, ensuring a stable weight distribution, is determined by existing data validation techniques. It's a numerical correction performed to maintain the stability of the weight distribution when validating the fitted spatiotemporal weights. This validation technique typically involves detecting bias in the training samples; if an overall shift occurs, adjustments are made. To eliminate deviations.

[0089] Step S2 then performs spatiotemporal bias correction to eliminate the offset caused by spatiotemporal conditions on the image. This is achieved by correcting each component, and the corresponding model formula is as follows:

[0090] ;

[0091] in,

[0092] : Corrected Each portrait component;

[0093] : No. before correction Each portrait component;

[0094] Spatiotemporal weights are used to perform multiplicative corrections on each component.

[0095] Step S2 concludes with the reconstruction of the corrected image. The corresponding reconstruction formula is as follows:

[0096] ;

[0097] in,

[0098] : The new image vector formed after spatiotemporal correction.

[0099] The innovation of step S3, the spatiotemporal correction stage, lies primarily in the integrated data correction process that combines joint spatiotemporal weight modeling, spatiotemporal bias correction, and post-correction portrait reconstruction. This overcomes the limitations of existing technologies that only perform simple normalization and lack weight correction for each component, making the spatiotemporal correction stage more adaptable to data under different spatiotemporal conditions. Specifically, this stage achieves intelligent data correction by assigning weights to each component that depend on spatiotemporal conditions. This correction method significantly reduces the interference of spatiotemporal conditions on the credibility of portrait data, providing an accurate and robust data foundation for subsequent recommendation effect evaluation.

[0100] S3. Use the spatiotemporal correction image vector to correct the sample weights and obtain the sample weights;

[0101] Step S3 performs sample weight correction on the spatiotemporal corrected image vector output in step S2 to adapt to the effect of different samples on the overall training process.

[0102] Step S3 first performs sample weight modeling. In order to measure the credibility of each sample in model training, it is necessary to perform weight modeling on the spatiotemporal correction profile. The weight depends on the degree of anomaly of the sample in the spatiotemporal correction profile. The degree of anomaly can be measured by Euclidean distance to measure the magnitude of the sample's deviation from the overall mean.

[0103] The corresponding model formula is:

[0104] ;

[0105] in,

[0106] :sample abnormal scores

[0107] : No. Spatiotemporal corrected image vector for each sample;

[0108] The equilibrium vector obtained from the overall mean of the spatiotemporally corrected image. In the formula The total number of samples.

[0109] Step S3 then performs sample weight correction, assigning weights to each sample based on the outlier score:

[0110] ;

[0111] in, Sample weights;

[0112] :sample The anomaly score is a quantitative measure of the difference between the profile vector obtained after spatiotemporal correction of each sample and the overall sample mean. Specifically, this process uses Euclidean distance for calculation; the Euclidean distance from each sample to the overall mean is the anomaly score for that sample.

[0113] The weight decay factor, a normal number, determines the rate at which the weight decreases as the outlier score increases. This factor is determined through statistical modeling of the outlier score for each sample in the spatiotemporally corrected profile sample. First, the difference between the spatiotemporally corrected profile vector and the overall mean is measured for each sample to obtain an outlier score sequence. Next, statistical analysis is performed on the outlier scores to calculate their mean and obtain a data center, while simultaneously measuring the dispersion of the distribution when outlier scores deviate significantly. Finally, the weight decay factor is adaptively set based on the dispersion of the outlier scores. This setting ensures that the weight decays stably and moderately as the outlier score increases, thereby suppressing the interference of outlier samples on the overall training process while avoiding unnecessary suppression of normal samples. This process achieves data adaptation, robustness, and stability in the weight decay setting, applicable to data distributions under different conditions, providing a more robust and reliable data foundation for the overall model.

[0114] Step S3 finally outputs the weight of each sample:

[0115] ;

[0116] And this forms the sample weights:

[0117] .

[0118] The innovation of step S3 lies primarily in the joint modeling of sample weights using spatiotemporally corrected profiles and sample anomaly scores. This joint modeling overcomes the problem of insufficient robustness caused by assigning uniform weights to samples and lacking differentiated modeling in existing technologies, providing a more robust and reliable data foundation for subsequent steps. Existing technologies typically assign approximately the same weight to each sample or set weights solely based on sample labels, lacking quantitative modeling of the actual data distribution information of the samples. This method, however, relies on spatiotemporally corrected profiles to measure anomaly scores for each sample, and then performs weight correction based on the anomaly scores, thereby significantly suppressing the adverse effects of anomaly samples on the overall model training. This sample weight correction method is applicable to more conditions and has stronger robustness, providing a more stable and reliable data foundation for subsequent steps.

[0119] S4. Use sample weights to perform joint evaluation and obtain the joint evaluation score;

[0120] Step S4 evaluates each sample using the combined sample weight information. The combined evaluation loop includes weight normalization, combined evaluation score modeling, and combined evaluation score output.

[0121] Step S4 first performs weight normalization to eliminate the interference of sample size on the joint evaluation score and normalize the weights. The calculation formula is:

[0122] ;

[0123] This process ensures that the total weight of each sample is 1, which is more conducive to stable and reliable data modeling during joint evaluation.

[0124] Step S4 then performs joint evaluation score modeling, and the formula for calculating the joint evaluation score is as follows:

[0125] ;

[0126] in,

[0127] Joint evaluation score: measures the risk of anomalies in the overall sample;

[0128] like The value is directly proportional to the degree of risk of anomalies in the overall sample. This joint evaluation method overcomes the shortcomings of existing technologies that treat each sample equally and are insufficient to characterize the overall risk.

[0129] Step S4 concludes with the joint evaluation score output, and the output value is... This involves normalizing and aggregating samples under weights, which more clearly depicts the risk of anomalies in the overall sample set.

[0130] Existing techniques typically use simple summation or averaging of each sample to obtain an overall anomaly score, lacking joint modeling of sample weight information, resulting in a lack of robustness in the overall evaluation. This method relies on spatiotemporally corrected profiles to measure the anomaly score of each sample, and then uses normalization modeling based on the sample weights to form a joint evaluation score, which more clearly characterizes the overall risk of anomalies. This joint evaluation approach significantly overcomes the insufficient robustness caused by the lack of weight modeling in existing techniques, providing a more stable and reliable data foundation for subsequent anomaly detectors.

[0131] S5. Normalize and quantify the joint evaluation scores to obtain the quantified output scores;

[0132] Step S5 normalizes and quantifies the joint evaluation score, providing a stable and reliable data foundation for evaluating the effectiveness of advertising recommendations.

[0133] To eliminate the interference of different anomaly risk domains among samples on the quantification results, the joint evaluation scores need to be normalized. Normalization is achieved through min-max normalization, a mature technique, including:

[0134] The maximum score among the joint evaluation scores is set as the upper limit, and the minimum score is set as the lower limit.

[0135] Normalize each sample by linearly normalizing it to a value between 0 and 1 based on the maximum and minimum scores.

[0136] After normalization, each sample is assigned a normalized joint evaluation score, with a stable and intuitive numerical distribution, making it suitable as a quantitative input.

[0137] Finally, the normalized joint evaluation score is quantified using the existing mature technology of linear mapping. This process assigns a quantified score to each sample, with the score range from 0 to 100. The higher the value, the greater the risk of the sample being anomaly. This quantified score provides an intuitive and reliable data foundation for evaluating the effectiveness of advertising recommendations, helps to measure the overall level of risk of recommendations being anomaly, and provides a quantitative reference for adjusting marketing strategies.

[0138] In advertising recommendations, anomaly risk refers to the risk of biased, ineffective, or even negative recommendations. This risk mainly manifests in the following situations:

[0139] Recommended content is seriously mismatched with user interests (leading to decreased click-through rates and conversion failures, thus wasting marketing resources);

[0140] Negative feedback to recommendations (such as being blocked by users, not being interested, or not being applicable) can interfere with the subsequent performance of the recommendation model.

[0141] Recommendations that negatively impact overall marketing objectives (such as leading to wasted marketing funds or decreased customer satisfaction) will diminish marketing effectiveness.

[0142] After quantifying the risk into numerical values ​​through the above steps S1-S6, a foundation is provided for measuring the magnitude of recommendation risk in the advertising recommendation loop, assisting technical personnel in model optimization, resource allocation, and marketing strategy adjustment.

[0143] Example 2:

[0144] Reference Figure 5 This is the second embodiment of the present invention. This embodiment is based on the advertising recommendation effectiveness evaluation system based on multi-dimensional user profiles described in Embodiment 1, and its functional modules include:

[0145] The user profile quantification module performs attribute quantification processing on the collected multi-dimensional user profile data, transforming discrete, continuous, and categorical information into a unified profile vector representation, forming a numerical user feature expression that can be used for subsequent calculations. This module standardizes and structures user profiles and is the foundation of the system's input information.

[0146] Spatiotemporal Correction Module: Utilizing the time and geographic location information of the samples, this module performs spatiotemporal weight correction on the user profile vector, adjusting the influence of profile features to eliminate biases caused by spatiotemporal heterogeneity and ensuring the applicability and accuracy of profile features under different time and space conditions. This module achieves the fusion of multi-dimensional profiles with spatiotemporal conditions, enhancing the dynamic adaptability of the profile.

[0147] Sample weight correction module: Based on the spatiotemporally corrected profile vector and the conditions under which samples occur, such as clicks and conversions, this module calculates and adjusts sample weights to reflect the relative importance and influence of samples in the overall recommendation performance evaluation. Through comprehensive weight correction, this module enhances the ability to identify anomalous samples and key behaviors, improving the robustness and accuracy of the evaluation.

[0148] Joint evaluation score calculation module: Combining sample weights and spatiotemporal correction profile features, as well as relevant occurrence conditions, it performs multi-factor joint evaluation and outputs a unified joint evaluation score.

[0149] The normalized quantitative output module normalizes the joint evaluation scores, standardizing the evaluation results from different samples and scenarios into a unified quantitative output score. This facilitates comparison and application across samples and time and space conditions. The quantitative scores output by this module can serve as the basis for subsequent advertising recommendation decisions or risk warnings, enhancing the system's practicality and promotional value.

[0150] Example 3:

[0151] This embodiment also provides a computer device applicable to an advertising recommendation effect evaluation system based on multi-dimensional user profiles, including a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the advertising recommendation effect evaluation system based on multi-dimensional user profiles as proposed in the above embodiment.

[0152] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements an advertising recommendation effect evaluation system based on multi-dimensional user profiles as proposed in the above embodiments.

[0153] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0154] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0155] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0156] More specific examples (a non-exhaustive list) of computer-readable media include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which programs can be printed, because programs can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0157] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0158] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. An advertising recommendation effectiveness evaluation system based on multi-dimensional user profiles, characterized in that, The system's functional modules include: a user profile quantification module, a spatiotemporal correction module, a sample weight correction module, a joint evaluation score calculation module, and a normalized quantification output module; The application of the system's functional modules includes the following steps: Based on the user profile quantization module, the collected user profiles are quantified to obtain profile vectors; The spatiotemporal correction module is used to perform spatiotemporal correction on the profile vector obtained by the user profile quantization module to obtain the spatiotemporally corrected profile vector. The sample weights are obtained by performing sample weight correction on the spatiotemporal correction image vector output by the spatiotemporal correction module. Based on the sample weights output by the sample weight correction module, a joint evaluation is performed using the joint evaluation score calculation module to obtain the joint evaluation score. The joint evaluation score output by the joint evaluation score calculation module is input into the normalization and quantification output module for normalization and quantification to obtain the quantification output score, which serves as the basis for measuring the magnitude of recommendation risk in the advertising recommendation loop. The higher the score, the higher the advertising recommendation risk. The spatiotemporal correction module first performs spatiotemporal weight modeling, then performs spatiotemporal deviation correction, and finally performs post-correction image reconstruction. In spatiotemporal weight modeling, a spatiotemporal weight is applied to each image component. The spatiotemporal weight is determined by a spatiotemporal factor, and the corresponding spatiotemporal weight is modeled as a spatiotemporal function, the formula of which is: ； in, Spatiotemporal weights; : Data collection time; : Data collection location; : Sigmoid activation function; The sensitivity of spatiotemporal weights to time; The sensitivity of spatiotemporal weights to location; Bias value, to ensure stable weight distribution; When performing spatiotemporal deviation correction, the spatiotemporal correction module eliminates the offset of the image caused by spatiotemporal conditions. This is achieved by correcting each component, and the corresponding model formula is as follows: ； in, : Corrected Each portrait component; : No. before correction Each portrait component; Spatiotemporal weights are applied to each component using multiplication correction. In the sample weight correction module, sample weight modeling is performed first, then sample weight correction is performed, and finally the weights are output. When performing sample weight modeling, the sample The abnormal score is: ； in, : No. Spatiotemporal corrected image vector for each sample; The equilibrium vector obtained from the overall mean of the spatiotemporally corrected image. In the formula The total number of samples; When performing sample weight correction, the sample weight correction module assigns weights to each sample based on the outlier score: ； in, Sample weights; :sample Abnormal scores; Weight decay factor.

2. The advertising recommendation effect evaluation system based on multi-dimensional user profiles according to claim 1, characterized in that, The input profile in the user profile quantization module includes: basic user information, user interest tags, user behavior data, and user geographic information. This profile information is quantized using One-Hot encoding, Word Embedding, and normalization techniques to form a profile vector. .

3. The advertising recommendation effect evaluation system based on multi-dimensional user profiles according to claim 1, characterized in that, When the spatiotemporal correction module reconstructs the corrected image, the corresponding reconstruction formula is: ； in, : The new image vector formed after spatiotemporal correction.

4. The advertising recommendation effect evaluation system based on multi-dimensional user profiles according to claim 1, characterized in that, The sample weight correction module ultimately outputs the weight of each sample: ； And this forms the sample weights: 。 5. The advertising recommendation effect evaluation system based on multi-dimensional user profiles according to claim 1, characterized in that, The joint evaluation score calculation module first performs weight normalization, then models the joint evaluation score to obtain the joint evaluation score. .

Images