Webpage putting method, electronic equipment and storage medium

Abstract

The embodiment of the invention provides a webpage release method, electronic equipment, a storage medium and a computer program product. The method comprises the following steps: constructing a heating medium trend data center consisting of a plurality of target keywords based on first data associated with a target webpage and a target search engine, public trend hotspot data of the target search engine and second data in a website to which an alternative putting webpage belongs; on the basis of the heat value index of each target keyword in the heating medium trend data center, correlation screening is conducted on the alternative putting webpages, putting webpages are obtained, and low-quality putting webpages can be preliminarily screened out; based on a preset content knowledge graph, generating delivery content of the delivery webpage; and then, based on the put content, putting the put webpage to the target search engine. According to the method, the quality of the released webpage is effectively improved, and the page release efficiency is improved by automatically generating the release content.

Description

Technical Field

[0001] This application relates to the field of computer technology, and particularly to a web page placement method, an electronic device, a storage medium, and a computer program product. Background Art

[0002] Web page placement refers to the technology of placing the pages of a website on a search engine or retrieval platform. To improve the quality of the placed web pages, the web page placement method can be improved in two aspects: optimizing the placement content and taking offline low-quality web pages. In the prior art, the generation of placement content mainly adopts the way of manual writing, resulting in unstable web page quality and being unfavorable for the timely maintenance of placed web pages. On the other hand, scientific decision-making and continuous optimization of the effect for incremental web page placement based on traffic are also an important link in web page placement. With the continuous iteration of search algorithms of search engines and the intensification of the competition environment, traditional web page placement strategies have defects such as high blindness and lagging response in the incremental content placement link. The web page placement methods in the prior art mainly focus on basic functional modules, including keyword mining, competitor backlink tracking, in-site health detection, and historical data visualization analysis, etc. In practical applications, the web page placement solutions in the prior art have the following defects: the quality of placement content is unstable; the accuracy of traffic prediction during web page maintenance mainly depends on the analysis ability of users, resulting in unsatisfactory web page placement effects.

[0003] It can be seen that the web page placement methods in the prior art still need to be improved. Summary of the Invention

[0004] Embodiments of this application provide a web page placement method, which can effectively improve the quality of web page placement and the web page placement effect.

[0005] Correspondingly, embodiments of this application also provide an electronic device, a storage medium, and a computer program product to ensure the implementation and application of the above web page placement method.

[0006] To solve the above problems, embodiments of this application disclose a web page placement method, and the method includes: Construct a hot medium trend data center composed of a number of target keywords based on the first data associated with the target web page and the target search engine, the public trend hot data of the target search engine, and the second data within the website to which the alternative placement web page belongs; Perform relevance screening on the alternative placement web pages based on the heat value index of each target keyword in the hot medium trend data center to obtain placement web pages; Generate placement content for the placement web pages based on a preset content knowledge graph; Place the placement web pages on the target search engine based on the placement content.

[0007] The embodiments of the present application also disclose a computer-readable storage medium, in which computer-executable instructions are stored, and when the computer-executable instructions are executed by a processor, they are used to implement the method as described in the embodiments of the present application.

[0008] The embodiments of the present application also disclose a computer program product, including a computer program / computer-executable instructions, and when the computer program / computer-executable instructions are executed by a processor in an electronic device, they implement the method as described in the embodiments of the present application.

[0009] Compared with the prior art, the embodiments of the present application have the following advantages: By constructing a hot media trend data center composed of several target keywords based on the first data associated with the target web page and the target search engine, the public trend hot data of the target search engine, and the second data within the website to which the alternative placement web page belongs, and then, based on the heat value index of each target keyword in the hot media trend data center, performing relevance screening on the alternative placement web page to obtain a placement web page, low-quality placement pages can be initially screened out. Then, based on a preset content knowledge graph, generate the placement content of the placement web page; not only effectively improves the quality of the placement web page, but also improves the page placement efficiency by automatically generating the placement content; after batch placement of web pages, by performing multi-scale traffic time series prediction on the placed placement web pages, obtaining the prediction results of the preset placement effect evaluation indicators matched by each placement web page, obtaining the performance of the future time interval of the placement page cluster in the search results, and combining historical data to perform fluctuation analysis on the prediction results. Then, based on the fluctuation analysis results, perform placement maintenance processing on the corresponding placement web pages, thereby further improving the quality of the placement web pages and improving the web page placement effect by enhancing the scientific nature of web page placement. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 is one of the step flowcharts of the web page placement method disclosed in the embodiments of the present application; Figure 2 is a schematic diagram of the construction principle of the hot media trend data center in the web page placement method disclosed in the embodiments of the present application; Figure 3 is a schematic diagram of the closed-loop link of the web page placement method disclosed in the embodiments of the present application; Figure 4 is the second step flowchart of the web page placement method disclosed in the embodiments of the present application; Figure 5 is a schematic diagram of the structure of an exemplary device provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0011] To make the above objects, features, and advantages of the present application more obvious and understandable, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0012] The web page placement method disclosed in the embodiments of the present application adopts a set of closed-loop process mechanisms, integrates industry data (such as competitor data), public trend hot data, and in-site new product data of the website, and with the help of artificial intelligence content production capabilities and time series prediction capabilities, improves the quality and efficiency of the placed web pages and the web page placement effect through incremental web page placement.

[0013] As Figure 1 shown, the web page placement method disclosed in the embodiments of the present application includes: step 102 to step 108.

[0014] Step 102, construct a hot media trend data center composed of several target keywords based on the first data associated with the target web page and the target search engine, the public trend hot data of the target search engine, and the second data within the website to which the alternative placement web page belongs.

[0015] Among them, the target web page is the web page of the industry-leading website of the website to which the alternative placement web page belongs and / or the web page of the competitor website; the target search engine is the search engine for placing the alternative placement web page; the first data includes: the hot topics and / or keywords of the target web page in the target search engine; the second data includes: in-site search keywords and in-site new product data.

[0016] Taking the Google search engine as an example of the target search engine, in the scenario of placing the web page of website A on the Google search engine, the target web page is the website page of the same-industry merchants of website A (such as the website page of the competitor). The hot topics and / or keywords of the website page of the same-industry merchants of website A placed on the Google search engine can be obtained by using existing technologies to obtain the first data. The in-site search keywords of website A and the keywords refined from the in-site new product titles can be obtained by using existing technologies.

[0017] Among them, the public trend hot data of the target search engine refers to data with high access frequency, short-term popularity, and crucial importance to system performance, and its characteristics include high access frequency and short-term popularity. Optionally, the trend hot data can be obtained through the Google Cloud public data set according to the search interest index, time trend, and industry classification. Among them, the search interest index is used to reflect the search volume ratio of keywords relative to the reference time or region (such as a standardized value from 0 to 100); the time trend is used to represent the search volume fluctuation displayed by hour, day, month, or custom period; the industry classification is used for comparative analysis of search interests based on industry categories (such as technology, entertainment).

[0018] The following will describe the specific implementation manners of this step in conjunction with Figure 2 the schematic diagram of the structure principle of the heat medium trend data center shown in

[0019] Optionally, constructing a heat medium trend data center consisting of a number of target keywords based on the first data associated with the target web page and the target search engine, the public trend hot data of the target search engine, and the second data within the website to which the alternative placement web page belongs, includes: refining keywords in the titles of new products within the website based on the in-site new product data; fusing the hot topics of the target web page in the target search engine and the keywords, the in-site search keywords, and the keywords extracted from the in-site new product data to obtain an original keyword pool; performing keyword cleaning processing on the original keyword pool to obtain seed words; performing keyword expansion processing on the seed words to obtain expanded keywords; merging the seed words and the expanded keywords to obtain a data set of keywords; using an artificial intelligence model to perform tagging and screening on the keywords in the data set based on the public trend hot data of the target search engine to obtain a number of target keywords; and constructing a heat medium trend data center based on the number of target keywords.

[0020] Among them, the in-site search keywords and the in-site new product data can be obtained by analyzing and processing the in-site offline data of the website to which the alternative placement web page belongs.

[0021] Optionally, when refining keywords in the titles of new products within the website based on the in-site new product data, one or more words such as effect description words, trend description words, and names can be extracted from the titles of new products within the website as keywords based on keyword extraction techniques in the prior art.

[0022] Then, the hot topics of the target web page in the target search engine and the keywords, the in-site search keywords, and the keywords extracted from the in-site new product data can be comprehensively used as the original keyword pool.

[0023] Optionally, the seed words are usually keywords related to the core scenario or theme, with moderate search volume and low competition degree, and have high relevance and expandability, and can effectively derive long-tail words. In the process of performing keyword cleaning processing on the original keyword pool to obtain seed words, one or more of the following keyword cleaning operations on the original keywords can be implemented by using natural language processing tools in the prior art and translation models in the prior art: deleting special characters, filtering out invalid words, filtering out keywords with too low search volume, filtering out keywords with too high competition degree or irrelevant to the specified scenario, performing word segmentation and truncation processing on long keywords, performing unified English translation on keywords, etc., so as to retain keywords highly relevant to the specified scenario and keywords with high conversion potential, thereby forming seed words.

[0024] Optionally, in the process of expanding the seed words to obtain expanded keywords, the product words in the seed words can be extracted through a named entity recognition model, and then an existing word expansion tool can be used for word expansion and cleaning to obtain result words as the expanded keywords. Through the natural language keyword search awareness recognition ability, semantic associations (such as synonyms, scenario words, etc.) and long-tail expansion requirements are realized, extending from the core words outward to construct a keyword matrix covering a wider range of requirements.

[0025] After keyword cleaning and word expansion processing, the seed words and the expanded keywords are merged to obtain a keyword dataset.

[0026] Next, the large language model can be used to combine public trend hot data to perform tagging and screening on the merged keyword dataset to obtain a number of target keywords. The target keywords include: hot scenario words, non-hot scenario words, hot product words, and non-hot product words.

[0027] For example, the large language model can be triggered through prompt words to identify potential hot scenario words, non-hot scenario words, hot product words, and non-hot product words from the keyword dataset obtained in the previous step. Taking the e-commerce scenario as an example, the scenario words represent keywords with clear transaction / cooperation intentions when users search, and are usually related to brands, services, solutions, or purchase decisions, such as service words and solution words; the product words represent keywords that focus on specific product attributes or models when users search, and are usually used to find the characteristics, parameters, or comparison information of a certain type of product, such as product words with attributes or parameters. Then, corresponding keyword category labels are set based on the recognition results.

[0028] On the other hand, calculate the heat value index of each keyword. The heat value index is a quantitative index that comprehensively evaluates the potential and optimization feasibility of keywords, and helps to screen high-cost performance keywords through multi-dimensional data fusion. For example, the heat value index can be calculated according to the search volume, competition degree, and click-through rate of the keywords.

[0029] In some alternative embodiments, the heat value index can be calculated using the following formula: HVI=(SearchVolume7d / CompScore)×log(ClickThroughRate); where SearchVolume7d represents the 7-day search volume of the keyword, CompScore represents the competition degree, and ClickThroughRate represents the click-through rate. The 7-day search volume, competition degree, and click-through rate of the keywords can be obtained by statistically analyzing historical data.

[0030] Finally, a heat medium trend data center is constructed from the filtered target keywords, the types and heat value indexes of each target keyword.

[0031] Step 104: Based on the heat value indexes of each target keyword in the heat medium trend data center, perform relevance screening on the alternative placement web pages to obtain placement web pages.

[0032] Optionally, the performing relevance screening on the alternative placement web pages based on the heat value indexes of each target keyword in the heat medium trend data center to obtain placement web pages includes: screening out alternative web page keywords based on the heat value indexes of each target keyword in the heat medium trend data center; performing product recall based on the alternative web page keywords to obtain the product features and relevance scores of the recalled products; performing duplicate removal processing on the web pages associated with the products based on the product similarity expressed by the product features to obtain alternative placement web pages; and selecting the alternative placement web pages as the placement web pages based on the matching result between the relevance scores and a preset combination ladder screening condition of relevance scores and the number of products.

[0033] In the embodiments of the present application, the higher the heat value index, the higher the value of the web page associated with the product. By calculating and sorting the heat value indexes of all keywords, and then dividing the interval according to the percentile, the screening threshold interval of the heat value index is obtained (for example, the top 20% is set as high value), and the keywords of the web pages to be placed can be screened out through the threshold of this index as the alternative web page keywords.

[0034] Optionally, by invoking the service provided by the target search engine, perform product recall based on the alternative web page keywords, so as to obtain the product features and relevance scores of the recalled products.

[0035] In some alternative embodiments, the target search engine uses the BERT model to establish a product feature vector index and can provide a service for recalling relevant products through keywords externally. On the other hand, the target search engine performs a relevance score on the recalled products and keywords through a similarity model and outputs the relevance scores of the products recalled through keywords. For example, first, based on the alternative web page keywords, invoke the product recall service provided by the target search engine to obtain the product identifiers of all products associated with the alternative web page keywords output by the service and the relevance scores of the products corresponding to each product identifier. Then, based on the product identifiers, further invoke the product completion service provided by the target search engine to obtain product title features associated with the products, etc., as the page features of the web pages associated with the products.

[0036] Next, encoding techniques can be used to vectorize each page feature to obtain the page feature vectors of each web page associated with the product. Then, a vector cosine similarity model can be used to calculate the web page similarity based on the page feature vectors, and duplicate removal can be performed on each web page associated with the product to obtain alternative placement web pages.

[0037] In the specific implementation process, each product in the search engine has its own product features and relevance scores to search keywords. The product features and relevance scores are in one-to-one correspondence with the product identifiers. Moreover, a keyword often may recall multiple products, that is, the product features recalled by a keyword may be multiple. Since products are recalled based on relevance when recalling products based on keywords, all the recalled product features are similar, only the similarity may vary, and the corresponding relevance scores may be different. Therefore, it is necessary to filter out the web pages associated with the product features with relatively low overall relevance scores through a threshold.

[0038] In the embodiments of the present application, stepwise combination conditions can be set in combination with the number of recalled products and the relevance scores, denoted as "stepwise screening conditions for the combination of relevance scores and product numbers". For example, setting 1 < count <= 10 && score >= 0.8 as the combination conditions for one step, and setting 10 < count <= 30 && 0.5 < score < 0.8 as the combination conditions for another step. By setting the stepwise screening conditions for the combination of relevance scores and product numbers, the relevance of page content and the richness of web page products are taken into account during web page placement, so as to obtain more web pages to be placed. That is, for web pages with relatively high relevance, the number of recalled products can be slightly reduced, and for web pages with relatively low relevance, the number of recalled products can be increased.

[0039] Step 106, generate the placement content of the placement web page based on a preset content knowledge graph.

[0040] After screening out the pages to be launched, that is, the placement web pages, through the foregoing steps, in order to obtain as much traffic as possible, next, the web page quality is improved through content optimization.

[0041] Optionally, the generating the placement content of the placement web page based on a preset content knowledge graph includes: obtaining the product features associated with the placement web page; using an artificial intelligence large model, based on the product features, and combining the classification industry general knowledge and hot media trend data associated with the product features in the preset content knowledge graph, to generate the placement content of the placement web page. Wherein, the placement content includes, but is not limited to, metadata tags and copywriting.

[0042] In the embodiment of the present application, in order to solve the labor cost and efficiency problems of batch content optimization and ensure the quality of content, a vector library of the content knowledge graph is established, in which the classification industry general knowledge, product feature data and hot media trend data are embedded. Then, the vector library of the content knowledge graph is plugged in through the retrieval augmented generation technology (RAG technology) using the artificial intelligence big model capability to realize the metadata tag generation and product short copy generation of the web page. Among them, the metadata tags include: title and page description.

[0043] For example, the product features of the recalled products can be pre-embedded into a vector database using vector encoding technology combined with pre-collected classified industry knowledge and hot media trend data; then, the vectors in the vector database are processed using a large model and NLP (Natural Language Processing) named entity technology to obtain a content knowledge graph, which is stored as a preset content knowledge graph. Optionally, the preset content knowledge graph can be retrieved based on product features.

[0044] In the application process, the existing technology can be used to first obtain the page data of each delivery web page, and the page data includes the page content. After that, refer to the method described above or use the existing technology to obtain the product features associated with each delivery web page. Then, the page content of each delivery web page and the product features associated with the delivery web page are used as input data to trigger the artificial intelligence model to combine the input data and the classified industry general knowledge, hot media trend data and other information associated with the product features in the content knowledge graph to automatically generate the target content such as the title, description and short copy of the corresponding web page, thereby obtaining the metadata tag and copy of the delivery web page as part of the delivery content.

[0045] Step 108: Based on the delivery content, deliver the delivery webpage to the target search engine.

[0046] Optionally, a sitemap generation tool is used to generate a delivery sitemap for batch delivery web pages based on the delivery content (such as metadata tags and copy), and finally the delivery of the batch of new delivery web pages is achieved through the sitemap submission application interface of the target search engine.

[0047] When a batch of web pages are placed in the background of the target search engine (such as Google), the target search engine will include the placed web pages, and then generate traffic. It is particularly important to scientifically analyze the traffic, exposure, average ranking and other trends of a certain batch of web page clusters through the relevant indicators provided by the search engine and the site access log. By analyzing these trends, we can find out which web pages in this batch have better drainage effects, which web pages have drainage potential, and which web pages have reduced drainage effects over time. For web pages with reduced drainage effects, corresponding strategies can be used to enhance the drainage effect or offline low-quality web pages.

[0048] In the embodiments of the present application, Figure 3 The integrated closed-loop chain shown, from web page content generation and delivery to page search result performance timing analysis, and then to web page offline and page content update, can further improve the scientific nature of web page delivery.

[0049] In some optional embodiments, such as Figure 4 As shown, after step 108 , the method further includes: step 110 , step 112 and step 114 .

[0050] Step 110, performing multi-scale traffic time series prediction on the delivered web pages, and obtaining prediction results of preset delivery effect evaluation indicators matched by each of the delivered web pages.

[0051] The traffic of web pages has time series characteristics, namely time dependence, periodicity and seasonality. However, the interference and noise of the time series characteristics of traffic (such as Google algorithm adjustment, major international events, etc.) have a large impact. It is difficult for traditional seasonal analysis time series models to fit exogenous variables. In order to ensure the accuracy of model analysis, in the embodiments of the present application, a large model time series model (such as timeGpt) is used based on the time series characteristics of traffic, combined with the fitted exogenous event variables, to train the large model time series model, and based on the trained large model time series model, the performance of the web page to be placed in the target search engine is predicted to obtain a prediction result. The prediction result is used to indicate whether the preset delivery effect evaluation index (such as exposure or average ranking) of the web page in the future is stable, for example, whether the preset delivery effect evaluation index is lower than the expected value (such as a decrease of 5% confidence).

[0052] Optionally, perform multi-scale traffic time series prediction on the delivered landing pages to obtain the prediction results of the preset delivery effect evaluation indicators matching each of the landing pages, including: obtaining the multi-scale time series features of the delivered landing pages in a preset time period; embedding the features of the preset exogenous variables into each of the multi-scale time series features to obtain the spliced features; and performing mapping processing based on the spliced features to obtain the prediction results of the preset delivery effect evaluation indicators matching the landing pages. Among them, the preset delivery effect evaluation indicators include, but are not limited to, any one of the following: web page performance indicators such as exposure volume, ranking, and unique visitor volume; the multi-scale time series features include one or more of the following: features in multiple dimensions such as user behavior time series features, search engine performance time series indicators, page performance (APM) time series indicators, and page content time series features. The multiple preset time periods can be, for example, three time periods: 30 days, 60 days, and 90 days.

[0053] Optionally, obtaining the multi-scale time series features of the delivered landing pages in a preset time period includes: obtaining the historical data of the delivered landing pages in the preset time period respectively; and extracting the multi-scale time series features of the delivered landing pages in each of the preset time periods based on the historical data. For example, extract the historical data of the delivered landing pages within 30 days, within 60 days, and within 90 days respectively, and then further extract the multi-scale time series features corresponding to each time period based on the historical data of these three time periods.

[0054] Among them, the multi-scale time series features include one or more of the following dimensions of features: search engine performance time series indicators, page performance (APM) time series indicators, page content time series features, etc.

[0055] Optionally, the search engine performance time series indicators include, but are not limited to, one or more of the following: exposure volume, click volume, average ranking, etc. The search engine performance time series indicators cannot be obtained through the search engine background. In the embodiments of the present application, the data channels between the search engine background and the cloud (Google background and Google Cloud) are used to import the page performance indicators and search result performance indicators of the web page into the cloud, and then the data in the cloud is synchronized to the time series feature library through a scheduled task.

[0056] Optionally, the page performance time series indicators are used to describe the page performance of the web page and include, but are not limited to, one or more of the following: first screen loading time, front-end page performance indicators (such as the duration of content drawing on the home page, the duration of the largest view drawing on the page, etc.). The method for obtaining the page performance time series indicators can be referred to the prior art and will not be elaborated in the embodiments of the present application.

[0057] Optionally, the page content time-series features are used to describe the features associated with the page content, including but not limited to those associated in the time dimension: product copywriting, keyword density of page metadata entities, content length, content depth, content freshness, etc. Among them, the keyword density of page metadata entities can be obtained by using TF-IDF (Term Frequency-Inverse Document Frequency) or BERT model encoding, and the content freshness can be obtained by the product of the content effective time decay factor, update intensity factor, and QDF (Query Deserves Freshness) weight. In some optional embodiments, the page content time-series features can be obtained by fusing web page content (mainly including titles, content relevance, metadata tag copywriting, etc.) and user behavior metrics (such as user page dwell time, web traffic), and according to the effective time of web page content delivery. For example, first, the web page content is structured to obtain the structured features of the web page content, and then, the structured features of the web page content are associated in the time dimension to obtain the page content time-series features.

[0058] Optionally, the user behavior metrics include: user page dwell time, web traffic, etc. For example, the access logs of the target search engine will record time-series metrics such as user dwell time and traffic. In the specific application process, the dwell time of users on the web page, the number of access users for each web page, etc. can be obtained through the service interface provided by the method of embedding points and sampling in the target search engine.

[0059] Optionally, the structured features of the web page content are associated in the time dimension to form a content-performance joint feature with timeliness as the page content time-series feature. For example, the structured features of the web page content are aligned with the time series metrics according to the effective timestamp (for example, if a certain web page copywriting goes online on October 1, 2024, then the corresponding structured features of the web page content are associated from that date).

[0060] Optionally, features of a preset exogenous variable are respectively embedded into each of the multi-scale time-series features to obtain the concatenated features corresponding to each time period, including: for each preset time period, the multi-scale time-series features and the features of the preset exogenous variable are concatenated to obtain the concatenated feature corresponding to the preset time period.

[0061] Optionally, performing a mapping process based on the concatenated features to obtain the prediction result of the preset delivery effect evaluation index matching the delivered web page, including: using the attention mechanism guided by the fitted exogenous variable to dynamically adjust the weights of each dimension of the multi-scale time-series features in the concatenated features, and using the dynamically adjusted weights to perform weighted fusion on the multi-scale time-series features to obtain a fused feature; performing feature mapping on the fused feature to obtain the prediction result of the preset delivery effect evaluation index matching the delivered web page.

[0062] Optionally, an attention mechanism guided by fitting exogenous variables is used to dynamically adjust the weights of each dimension of the multi-scale time series features in the splicing features, including: introducing the features of the preset exogenous variables as conditional information into the attention layer of the preset large model time series model, and dynamically adjusting the weights of each feature dimension of the multi-scale time series features in the splicing features based on the conditional information. Among them, the preset exogenous variables include variables of one or more of the following events: algorithm update events, periodic / seasonal events, holidays, etc.

[0063] Optionally, the large model time series model can adopt the TimeGPT-PatchTST architecture. After obtaining the multi-scale time series features of the delivered web page clusters in a preset time period based on historical data, the large model time series model uses parallel multi-scale convolution to perform feature encoding on the multi-scale time series features to obtain a first feature encoding vector. On the other hand, the features of the preset exogenous variables are encoded to obtain a second feature encoding vector. Then, the first feature encoding vector and the second feature encoding vector are spliced and used as the input of the fully connected layer of the large model time series model to realize embedding and fitting of exogenous variables, and the fully connected layer generates the weights of each dimension feature in the multi-scale time series features based on the input.

[0064] Among them, the fitting exogenous variables are determined according to the specific prediction scenario. For example, when predicting the web page exposure volume in the next 30 days, the fitting exogenous variables can be determined according to the predictable exogenous events within the next 30 days.

[0065] Finally, based on the weights of each dimension feature, weighted fusion is performed on each dimension feature in the multi-scale time series features to obtain a fusion feature.

[0066] After that, the output layer of the large model time series model performs feature mapping processing on the fusion feature to obtain the prediction result of the preset delivery effect evaluation index. For example, obtain the prediction result of the exposure volume, or obtain the prediction result of the average ranking, etc.

[0067] By introducing fitting exogenous variables, the accuracy of the prediction result can be improved.

[0068] In some alternative embodiments, a multi-task prediction method can be adopted to call the large model time series model to predict the prediction results for preset time periods (such as the next 30, 60, and 90 days).

[0069] In the embodiments of the present application, the large model time series model is obtained by fitting exogenous event variables using the multi-scale time series features of the web page clusters after delivery and then fine-tuning. Before fine-tuning the large model time series model, first, referring to the method for obtaining the multi-scale time series features described above, the multi-scale time series features corresponding to several time points are obtained, and further preprocessing such as missing value filling and normalization is performed on the multi-scale time series features to obtain input features that meet the input requirements of the large model time series model. On the other hand, the events associated with the multi-scale time series features corresponding to each time point are obtained, and the features of the exogenous variables corresponding to the events are obtained. Then, the multi-scale time series features and the features of the exogenous variables corresponding to each time point are concatenated to obtain the concatenated features corresponding to the corresponding time points, which are used as training samples. Then, the large model time series model is fine-tuned based on the training samples.

[0070] In the process of fine-tuning the large model time series model based on the training samples, the large model time series model uses parallel multi-scale convolution to perform feature encoding on the multi-scale time series features to obtain a first feature encoding vector. On the other hand, the features of the exogenous variables are encoded to obtain a second feature encoding vector. Then, the first feature encoding vector and the second feature encoding vector are concatenated as the input of the fully connected layer of the large model time series model to realize the embedding of the exogenous variables, and the fully connected layer uses the input second feature encoding vector as conditional information to dynamically generate the weights of the features of each dimension of the multi-scale time series features in the input first feature encoding vector. Then, through the attention mechanism layer of the large model time series model, the features of each dimension in the multi-scale time series features are weighted and fused based on the weights of the features of each dimension to obtain a fused feature. Finally, the fused feature is subjected to feature mapping processing through the output layer of the large model time series model to obtain the prediction result of the preset delivery effect evaluation index.

[0071] Next, fluctuation analysis is performed on the preset result to verify the reliability of the prediction result. For example, the prediction result can be fitted with the historical performance of the web page for fluctuation analysis, and then the fluctuation confidence interval is verified through ADF (Augmented Dickey-Fuller Test, a statistical test method for judging whether time series data has stationarity). If the prediction result satisfies the fluctuation confidence interval, the loss of the large model time series model is calculated based on the prediction result and the true value between the preset delivery effect evaluation indexes of the web page, and the model parameters are optimized with the goal of reducing the loss, and the large model time series model is iteratively fine-tuned.

[0072] For the specific implementation of the fluctuation analysis of the preset result, refer to the prior art, and it will not be elaborated in the embodiments of the present application.

[0073] Step 112: Conduct a volatility analysis on the prediction result in combination with historical data to obtain the volatility analysis result of the preset delivery effect evaluation index.

[0074] After predicting the prediction result for the preset time period, it is first necessary to conduct a volatility analysis on the prediction result to determine whether the volatility range of the prediction result is within the preset confidence interval. Optionally, the ADF test method can be used to test the stationarity of the prediction result for the preset time period to determine whether the volatility of the prediction result relative to historical data is within the confidence interval. If the volatility of the prediction result relative to historical data is within the confidence interval, then further combine the historical data and the prediction result within the prediction time period to obtain the volatility analysis result of the predicted preset delivery effect evaluation index. Among them, the volatility analysis result includes: the volatility analysis result indicating that the preset delivery effect evaluation index decreases and the volatility is within the preset confidence interval, and the volatility analysis result indicating that the preset delivery effect evaluation index increases and the volatility is within the confidence interval. If the volatility of the prediction result relative to historical data is outside the confidence interval, it indicates that there may be a problem with the accuracy of the prediction result, and re-prediction can be performed.

[0075] For the specific implementation of conducting a volatility analysis on the prediction result using the ADF test method, refer to the prior art and will not be elaborated in the embodiments of the present application.

[0076] For example, for each delivered web page, when the prediction result is the ranking, it is possible to first analyze whether the predicted ranking within the preset time period is stable through ADF. If it is stable, then further combine the historical ranking and the predicted ranking to obtain the ranking volatility of each delivered web page, and obtain one or more of the following volatility analysis results: the volatility analysis result indicating that the ranking decreases and the volatility is within the preset confidence interval, and the volatility analysis result indicating that the ranking increases and the volatility is within the preset confidence interval.

[0077] For another example, for each delivered web page, when the prediction result is the exposure volume, it is possible to first analyze whether the predicted exposure volume within the preset time period is stable through ADF. If it is stable, then further combine the historical exposure volume and the predicted exposure volume to obtain the exposure volume volatility of each delivered web page, and obtain one or more of the following volatility analysis results: the volatility analysis result indicating that the exposure volume decreases and the volatility is within the preset confidence interval, and the volatility analysis result indicating that the exposure volume increases and the volatility is within the preset confidence interval.

[0078] After obtaining the volatility analysis results of the preset delivery effect evaluation index for each web page, further perform delivery maintenance processing on the delivered web pages based on the maintenance strategy of the delivered web pages according to the volatility analysis results.

[0079] Step 114: Based on the fluctuation analysis result, perform delivery maintenance processing on the corresponding delivery web page, where the delivery maintenance processing includes: web page offline processing or web page update processing.

[0080] As described above, the fluctuation analysis result of the delivered web page is used to indicate the fluctuation of the preset delivery effect evaluation index of the web page in the preset time period. Next, according to the fluctuation situation of the preset delivery effect evaluation index indicated by the fluctuation analysis result, perform delivery maintenance processing on the delivered web page.

[0081] Optionally, the performing delivery maintenance processing on the corresponding delivery web page based on the fluctuation analysis result includes: when the preset delivery effect evaluation index includes: ranking, and the fluctuation analysis result indicates that the preset delivery effect evaluation index decreases and the fluctuation is within the preset confidence interval, generating the delivery content of the delivery web page based on the preset content knowledge graph, and updating the delivery web page based on the delivery content; or, when the preset delivery effect evaluation index includes: exposure volume, and the fluctuation analysis result indicates that the preset delivery effect evaluation index decreases and the fluctuation is within the preset confidence interval, perform offline processing on the delivery web page with the exposure volume less than the preset threshold.

[0082] For example, when the preset delivery effect evaluation index is ranking, if the fluctuation analysis result indicates that the ranking (i.e., the preset delivery effect evaluation index) decreases and the fluctuation is within the preset confidence interval, that is, the ranking fluctuation interval of the delivery web page decreases within the confidence interval during the predicted time interval, then it can jump to the step of generating the delivery content of the delivery web page based on the preset content knowledge graph, and update the retrieval information and page content of the delivery web page in the target search engine based on the generated delivery content, so as to update the delivery web page. Among them, the specific implementation manner of updating the delivery web page based on the delivery content refers to the prior art and will not be elaborated in the embodiments of the present application. By updating the web page content in a timely manner according to the fluctuation analysis result, the traffic loss caused by the decrease in the web page ranking can be avoided.

[0083] Another example, when the preset delivery effect evaluation index is exposure volume, if the fluctuation analysis result indicates that the exposure volume of a certain delivered web page decreases and the fluctuation is within the preset confidence interval, then the web page without exposure can be taken offline. The specific implementation manner of performing offline processing on the delivery web page with the exposure volume less than the preset threshold refers to the prior art and will not be elaborated in the embodiments of the present application. By performing web page offline processing according to the fluctuation analysis result, the low-quality delivered web pages can be reduced.

[0084] Optionally, multiple prediction tasks can also be adopted to respectively perform multi-scale traffic time series prediction on the delivered placement web pages, and obtain the prediction results of various preset placement effect evaluation indicators matched with each placement web page; then, perform fluctuation analysis on each of the prediction results to obtain the fluctuation analysis results of multiple preset placement effect evaluation indicators; finally, combine the fluctuation analysis results of multiple preset placement effect evaluation indicators, and perform placement maintenance processing on the corresponding placement web pages.

[0085] In summary, in the web page placement method disclosed in the embodiments of the present application, a hot media trend data center composed of several target keywords is constructed based on the first data associated with the target web page and the target search engine, the public trend hot data of the target search engine, and the second data within the website to which the alternative placement web page belongs; based on the heat value index of each target keyword in the hot media trend data center, perform relevance screening on the alternative placement web pages to obtain placement web pages, which can initially screen out low-quality placement pages; generate the placement content of the placement web pages based on a preset content knowledge graph; then, based on the placement content, deliver the placement web pages to the target search engine, which not only effectively improves the quality of the placement web pages, but also improves the page placement efficiency by automatically generating the placement content.

[0086] Further, after batch placement of web pages, perform multi-scale traffic time series prediction on the delivered placement web pages to obtain the prediction results of the preset placement effect evaluation indicators matched with each placement web page, obtain the performance of the future time interval of the placement page cluster in the search results, and perform fluctuation analysis on the prediction results in combination with historical data. Then, based on the fluctuation analysis results, perform placement maintenance processing on the corresponding placement web pages to improve the scientific nature of web page placement, thereby further improving the quality of the placement web pages and improving the web page placement effect.

[0087] The web page placement method disclosed in the embodiments of the present application forms an integrated closed-loop link from web page content generation, placement to the analysis of the time series performance of page search results, then to web page offline and page content update, improving the scientific nature of web page placement decisions.

[0088] At the implementation level, by combining a time series prediction model and search engine optimization effect evaluation, this method uses a large model time series model for feature modeling, integrates exogenous variable analysis, and constructs a dynamic decision-making mechanism based on the ADF test, enabling the web page update strategy to have confidence support with statistical significance, thereby improving the scientific nature of web page placement decisions.

[0089] Based on the above embodiments, this embodiment further provides a web page placement device, and the device includes: A heat medium trend data center construction module is used to construct a heat medium trend data center composed of several target keywords based on the first data associated with the target web page and the target search engine, the public trend hot data of the target search engine, and the second data within the website to which the alternative placement web page belongs; A placement web page screening module is used to perform relevance screening on the alternative placement web pages based on the heat value index of each target keyword in the heat medium trend data center to obtain placement web pages; A placement content generation module is used to generate the placement content of the placement web page based on a preset content knowledge graph; A web page placement module is used to place the placement web page on the target search engine based on the placement content.

[0090] Optionally, after placing the placement web page on the target search engine based on the placement content, the device further includes: A placement effect prediction module is used to perform multi-scale traffic time series prediction on the placed placement web pages to obtain the prediction results of the preset placement effect evaluation indicators matched by each placement web page; A prediction result fluctuation analysis module is used to perform fluctuation analysis on the prediction results in combination with historical data to obtain the fluctuation analysis results of the preset placement effect evaluation indicators; A placement web page maintenance module is used to perform placement maintenance processing on the corresponding placement web pages based on the fluctuation analysis results, where the placement maintenance processing includes: web page offline processing or web page update processing.

[0091] Optionally, the first data includes: the hot topics and / or keywords of the target web page in the target search engine; the second data includes: in-site search keywords and in-site new product data; constructing a heat medium trend data center composed of several target keywords based on the first data associated with the target web page and the target search engine, the public trend hot data of the target search engine, and the second data within the website to which the alternative placement web page belongs includes: Refining the keywords in the in-site new product titles based on the in-site new product data; Fusing the hot topics of the target web page in the target search engine and the keywords, the in-site search keywords, and the keywords extracted from the in-site new product data to obtain an original keyword pool; Performing keyword cleaning processing on the original keyword pool to obtain seed words; Performing keyword expansion processing on the seed words to obtain expanded keywords; Merging the seed words and the expanded keywords to obtain a data set of keywords; Using an artificial intelligence model, based on the public trend hot data of the target search engine, label and screen the keywords in the dataset to obtain a number of target keywords; Construct a heat medium trend data center based on the number of target keywords.

[0092] Optionally, the target keywords include: hot scenario words, non-hot scenario words, hot product words, and non-hot product words.

[0093] Optionally, the relevance screening of the alternative placement web pages based on the heat value index of each target keyword in the heat medium trend data center to obtain placement web pages includes: Based on the heat value index of each target keyword in the heat medium trend data center, screen and obtain alternative web page keywords; Based on the alternative web page keywords, perform product recall to obtain the product features and relevance scores of the recalled products; Based on the product similarity expressed by the product features, perform deduplication processing on the web pages associated with the products to obtain alternative placement web pages; Based on the matching result of the relevance score and the preset relevance score and product quantity combination ladder screening conditions, select the alternative placement web page as the placement web page.

[0094] Optionally, generating the placement content of the placement web page based on the preset content knowledge graph includes: Obtain the product features associated with the placement web page; Using an artificial intelligence large model, based on the product features, combined with the classification industry general knowledge and heat medium trend data associated with the product features in the preset content knowledge graph, generate the placement content of the placement web page.

[0095] Optionally, performing multi-scale traffic time series prediction on the placed placement web pages to obtain the prediction results of the preset placement effect evaluation indicators matched by each placement web page, including: Obtain the multi-scale time series features of the placed placement web pages in a preset time period; Embed the features of the preset exogenous variables into each of the multi-scale time series features to obtain splicing features; Based on the splicing features, perform mapping processing to obtain the prediction results of the preset placement effect evaluation indicators matched by the placement web pages.

[0096] Optionally, the mapping processing based on the splicing features to obtain the prediction results of the preset placement effect evaluation indicators matched by the placement web pages includes: Adopting an attention mechanism guided by fitting exogenous variables to dynamically adjust the weights of each dimension of the multi-scale time series features in the splicing features, and using the dynamically adjusted weights to perform weighted fusion on the multi-scale time series features to obtain fused features; Feature mapping is performed on the fusion features to obtain a prediction result of a preset delivery effect evaluation index matched by the delivery web page.

[0097] Optionally, performing delivery maintenance processing on the corresponding delivery webpage based on the fluctuation analysis result includes: In the case where the preset delivery effect evaluation index includes a ranking, and the fluctuation analysis result indicates that the preset delivery effect evaluation index decreases and the fluctuation is within a preset confidence interval, based on a preset content knowledge graph, the delivery content of the delivery webpage is generated, and the delivery webpage is updated based on the delivery content; or When the preset delivery effect evaluation index includes exposure, and the fluctuation analysis result indicates that the preset delivery effect evaluation index decreases and the fluctuation is within a preset confidence interval, offline processing is performed on the delivery web page whose exposure is less than a preset threshold.

[0098] The web page delivery device disclosed in the embodiment of the present application is used to implement the above-mentioned web page delivery method. The specific implementation methods of each module of the device refer to the specific implementation methods of the corresponding steps in the above-mentioned method embodiment, which will not be repeated here.

[0099] In summary, the web page delivery device disclosed in the example of the present application constructs a hot media trend data center composed of several target keywords based on the first data associated with the target web page and the target search engine, the public trend hot spot data of the target search engine, and the second data in the website to which the candidate delivery web page belongs; based on the heat value index of each of the target keywords in the hot media trend data center, the candidate delivery web pages are screened for relevance to obtain delivery web pages, and low-quality delivery pages can be preliminarily screened out; based on a preset content knowledge graph, the delivery content of the delivery web page is generated; and then, based on the delivery content, the delivery web page is delivered to the target search engine, which not only effectively improves the quality of the delivery web page, but also improves the page delivery efficiency by automatically generating delivery content.

[0100] Furthermore, after the web pages are delivered in batches, multi-scale traffic time series prediction is performed on the delivered web pages to obtain the prediction results of the preset delivery effect evaluation indicators matched by each delivered web page, and the performance of the delivery page cluster in the search results in the future time interval is obtained, and the prediction results are analyzed for fluctuations in combination with historical data. Thereafter, delivery maintenance processing is performed on the corresponding delivery web pages based on the fluctuation analysis results, thereby improving the scientific nature of web page delivery, thereby further improving the quality of the delivered web pages and improving the web page delivery effect.

[0101] The web page delivery device disclosed in the embodiments of the present application forms an integrated closed-loop link from web page content generation, timing analysis of the performance of page search results after delivery, to web page offline and page content update, improving the scientific nature of web page delivery decisions.

[0102] The embodiments of the present application also provide a non-volatile readable storage medium, in which one or more modules (programs) are stored. When the one or more modules are applied to a device, the device can be caused to execute instructions (instructions) for each method step in the embodiments of the present application.

[0103] The embodiments of the present application also provide a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the method as described in the embodiments of the present application.

[0104] The embodiments of the present application also provide an electronic device, including: a processor, and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the method as described in the embodiments of the present application. In the embodiments of the present application, the electronic device includes devices such as servers and terminal devices.

[0105] The embodiments of the present application also disclose a computer program product, including a computer program / computer-executable instructions. When the computer program / computer-executable instructions are executed by a processor in an electronic device, they implement the method as described in the embodiments of the present application.

[0106] The embodiments of the present disclosure can be implemented as a device configured as desired using any suitable hardware, firmware, software, or any combination thereof. The device may include electronic devices such as servers (clusters) and terminals. Figure 5 Exemplary device 500 that can be used to implement the various embodiments described in the present application is schematically shown.

[0107] For one embodiment, Figure 5 Exemplary device 500 is shown, which has one or more processors 502, a control module (chipset) 504 coupled to at least one of the (one or more) processors 502, a memory 506 coupled to the control module 504, a non-volatile memory (NVM) / storage device 508 coupled to the control module 504, one or more input / output devices 510 coupled to the control module 504, and a network interface 512 coupled to the control module 504.

[0108] The processor 502 may include one or more single-core or multi-core processors, and the processor 502 may include any combination of general-purpose processors or dedicated processors (such as graphics processors, application processors, baseband processors, etc.). In some embodiments, the device 500 is capable of serving as devices such as the server, terminal, etc. described in the embodiments of the present application.

[0109] In some embodiments, the device 500 may include one or more computer-readable media (such as the memory 506 or the NVM / storage device 508) having instructions 514, and one or more processors 502 combined with the one or more computer-readable media and configured to execute the instructions 514 to implement modules so as to perform the actions described in the present disclosure.

[0110] For one embodiment, the control module 504 may include any suitable interface controller to provide any suitable interface to at least one of the (one or more) processors 502 and / or any suitable device or component communicating with the control module 504.

[0111] The control module 504 may include a memory controller module to provide an interface to the memory 506. The memory controller module may be a hardware module, a software module, and / or a firmware module.

[0112] The memory 506 may be used, for example, to load and store data and / or instructions 514 for the device 500. For one embodiment, the memory 506 may include any suitable volatile memory, such as suitable DRAM. In some embodiments, the memory 506 may include double data rate type four synchronous dynamic random access memory (DDR4 SDRAM).

[0113] For one embodiment, the control module 504 may include one or more input / output controllers to provide an interface to the NVM / storage device 508 and the (one or more) input / output devices 510.

[0114] For example, the NVM / storage device 508 may be used to store data and / or instructions 514. The NVM / storage device 508 may include any suitable non-volatile memory (such as flash memory) and / or may include any suitable (one or more) non-volatile storage devices (such as one or more hard disk drives (HDDs), one or more compact discs (CD) drives, and / or one or more digital versatile discs (DVD) drives).

[0115] The NVM / storage device 508 may include storage resources that are part of the device on which the device 500 is installed, or it may be accessible by the device without being part of the device. For example, the NVM / storage device 508 may be accessed via the (one or more) input / output devices 510 through a network.

[0116] (One or more) input / output devices 510 can provide an interface for the apparatus 500 to communicate with any other suitable devices. The input / output devices 510 can include communication components, audio components, sensor components, etc. The network interface 512 can provide an interface for the apparatus 500 to communicate through one or more networks. The apparatus 500 can wirelessly communicate with one or more components of a wireless network according to any standard and / or protocol among one or more wireless network standards and / or protocols. For example, it can access a wireless network based on a communication standard, such as Bluetooth, WiFi, 2G, 3G, 4G, 5G, etc., or a combination thereof for wireless communication.

[0117] For one embodiment, at least one of the (one or more) processors 502 can be logically packaged together with one or more controllers (e.g., a memory controller module) of the control module 504. For one embodiment, at least one of the (one or more) processors 502 can be logically packaged together with one or more controllers of the control module 504 to form a system-in-package (SiP). For one embodiment, at least one of the (one or more) processors 502 can be logically integrated on the same die with one or more controllers of the control module 504. For one embodiment, at least one of the (one or more) processors 502 can be logically integrated on the same die with one or more controllers of the control module 504 to form a system-on-chip (SoC).

[0118] In various embodiments, the apparatus 500 can be, but is not limited to: a server, a desktop computing device, or a mobile computing device (e.g., a laptop computing device, a handheld computing device, a tablet computer, a netbook, etc.), etc. terminal devices. In various embodiments, the apparatus 500 can have more or fewer components and / or a different architecture. For example, in some embodiments, the apparatus 500 includes one or more cameras, a keyboard, a liquid crystal display (LCD) screen (including a touch screen display), a non-volatile memory port, multiple antennas, a graphics chip, an application specific integrated circuit (ASIC), and a speaker.

[0119] Among them, a main control chip can be used as the processor or the control module in the detection device. Sensor data, location information, etc. are stored in the memory or NVM / storage device. The sensor group can be used as an input / output device, and the communication interface can include a network interface.

[0120] Embodiments of the present application further provide an electronic device, including: a processor; and a memory storing executable code thereon, which, when executed, causes the processor to execute one or more of the methods as described in the embodiments of the present application. In the embodiments of the present application, various data can be stored in the memory, such as target files, file-application association data, and other various data, and can also include user behavior data, etc., thereby providing a data basis for various processes.

[0121] Embodiments of the present application further provide one or more machine-readable media storing executable code thereon, which, when executed, causes a processor to execute one or more of the methods as described in the embodiments of the present application.

[0122] For the apparatus embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and for the relevant parts, reference can be made to the partial description of the method embodiments.

[0123] The various embodiments in this specification are all described in a progressive manner. Each embodiment focuses on the differences from other embodiments, and for the same or similar parts among the various embodiments, reference can be made to each other.

[0124] The embodiments of the present application are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, and the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing terminal devices generate a device for implementing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0125] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing terminal device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device that implements the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0126] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device, so that a series of operation steps are executed on the computer or other programmable terminal device to generate a computer-implemented process. Thus, the instructions executed on the computer or other programmable terminal device provide steps for implementing the functions specified in one process or multiple processes and / or blocks. Figure 1 One process or multiple processes and / or blocks Figure 1 Steps for implementing the functions specified in one block or multiple blocks.

[0127] Although the preferred embodiments of the embodiments of the present application have been described, those skilled in the art can make additional changes and modifications once they learn the basic creative concepts. Therefore, the appended claims are intended to be construed as including the preferred embodiments and all changes and modifications falling within the scope of the embodiments of the present application.

[0128] Finally, it should also be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or terminal device comprising a series of elements not only includes those elements, but also includes other elements not expressly listed, or elements inherent to such process, method, article or terminal device. Without further limitation, an element defined by the statement "comprising one..." does not exclude the presence of another identical element in the process, method, article or terminal device comprising the said element.

[0129] The above has introduced in detail a web page delivery method, an electronic device, a storage medium and a computer program product provided by the present application. Specific examples are used in this article to elaborate on the principle and implementation manner of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present application.

Claims

1. A web page delivery method, characterized in that, The method includes: Constructing a hot media trend data center composed of a number of target keywords based on the first data associated with the target web page and the target search engine, the public trend hot data of the target search engine, and the second data within the website to which the alternative placement web page belongs; Performing relevance screening on the alternative placement web page based on the heat value index of each target keyword in the hot media trend data center to obtain a placement web page; Generating placement content for the placement web page based on a preset content knowledge graph; Based on the placement content, placing the placement web page on the target search engine.

2. The method according to claim 1, wherein After placing the placement web page on the target search engine based on the placement content, it further includes: Performing multi-scale traffic time series prediction on the placed placement web page to obtain the prediction results of the preset placement effect evaluation indicators matched by each placement web page; Performing fluctuation analysis on the prediction results in combination with historical data to obtain the fluctuation analysis results of the preset placement effect evaluation indicators; Based on the fluctuation analysis results, performing placement maintenance processing on the corresponding placement web page, where the placement maintenance processing includes: web page offline processing or web page update processing.

3. The method according to claim 1, wherein The first data includes: the hot topics and / or keywords of the target web page in the target search engine; the second data includes: in-site search keywords and in-site new product data; constructing a hot media trend data center composed of a number of target keywords based on the first data associated with the target web page and the target search engine, the public trend hot data of the target search engine, and the second data within the website to which the alternative placement web page belongs, includes: Refining the keywords in the in-site new product titles based on the in-site new product data; Fusing the hot topics of the target web page in the target search engine and the keywords, the in-site search keywords, and the keywords extracted from the in-site new product data to obtain an original keyword pool; Performing keyword cleaning processing on the original keyword pool to obtain seed words; Performing word expansion processing on the seed words to obtain expanded keywords; Merging the seed words and the expanded keywords to obtain a keyword data set; Using an artificial intelligence model to perform tagging and screening on the keywords in the data set based on the public trend hot data of the target search engine to obtain a number of target keywords; Constructing a hot media trend data center based on the number of target keywords.

4. The method according to claim 3, wherein The target keywords include: hot scenario words, non-hot scenario words, hot product words, and non-hot product words.

5. The method according to claim 1, characterized in that, Performing relevance screening on the alternative placement web page based on the heat value index of each target keyword in the hot media trend data center to obtain a placement web page, includes: Screening to obtain alternative web page keywords based on the heat value index of each target keyword in the hot media trend data center; Performing product recall based on the alternative web page keywords to obtain the product features and relevance scores of the recalled products; Performing duplicate removal processing on the web pages associated with the products based on the product similarity expressed by the product features to obtain alternative placement web pages; Based on the matching result of the correlation score with the preset correlation score and the product quantity combination ladder screening condition, select the alternative placement web page as the placement web page.

6. The method according to claim 1, characterized in that, Generating the placement content of the placement web page based on the preset content knowledge graph includes: Obtain the product features associated with the placement web page; Using an artificial intelligence large model, based on the product features, and combining the classification industry general knowledge and hot media trend data associated with the product features in the preset content knowledge graph, generate the placement content of the placement web page.

7. The method according to claim 2, characterized in that, Performing multi-scale traffic time series prediction on the placed placement web page to obtain the prediction results of the preset placement effect evaluation indicators matched by each placement web page, including: Obtain the multi-scale time series features of the placed placement web page in a preset time period; Embed the features of the preset exogenous variables into each of the multi-scale time series features to obtain spliced features; Based on the spliced features, perform mapping processing to obtain the prediction results of the preset placement effect evaluation indicators matched by the placement web page.

8. The method according to claim 7, wherein The mapping processing based on the spliced features to obtain the prediction results of the preset placement effect evaluation indicators matched by the placement web page includes: Adopt an attention mechanism guided by fitting exogenous variables to dynamically adjust the weights of each dimension of the multi-scale time series features in the spliced features, and use the dynamically adjusted weights to perform weighted fusion on the multi-scale time series features to obtain a fused feature; Perform feature mapping on the fused feature to obtain the prediction results of the preset placement effect evaluation indicators matched by the placement web page.

9. The method according to claim 2, wherein Based on the fluctuation analysis result, perform placement maintenance processing on the corresponding placement web page, including: When the preset placement effect evaluation indicator includes: ranking, and the fluctuation analysis result indicates that the preset placement effect evaluation indicator decreases and the fluctuation is within the preset confidence interval, generate the placement content of the placement web page based on the preset content knowledge graph, and update the placement web page based on the placement content; or, When the preset placement effect evaluation indicator includes: exposure volume, and the fluctuation analysis result indicates that the preset placement effect evaluation indicator decreases and the fluctuation is within the preset confidence interval, perform an offline process on the placement web page with an exposure volume less than the preset threshold.

10. An electronic device, characterized in that, Includes: A processor and a memory communicatively connected to the processor; The memory stores computer execution instructions; The processor executes the computer execution instructions stored in the memory to implement the method according to any one of claims 1-9.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer execution instructions, and when the computer execution instructions are executed by a processor, they are used to implement the method according to any one of claims 1-9.

12. A computer program product, comprising a computer program / computer-executable instructions, characterized in that, When the computer program / computer executable instructions are executed by a processor in an electronic device, the method according to any one of claims 1-9 is implemented.

Images