Urban advertisement putting management system and method

Abstract

The invention relates to the technical field of advertisement management, and discloses a city advertisement putting management system and method, and the method comprises the steps: obtaining a time queue of advertisement effect indexes, a time queue of external environment variables, and a time queue of competitive intelligence in a preset time window; performing time sequence coding on advertisement effect indexes, external environment variables (such as air temperature, humidity and holiday flag bits) and competitive intelligence (such as competitive product advertisement putting intensity within 500 meters) by using a double-layer LSTM network, and capturing intra-day and cross-day periodic mode characteristics; and feature representation is enhanced through a ternary attention network, and multi-source time sequence data are fused to generate an accurate advertisement putting prediction result. And the dynamic decision module based on the Softmax function further gives an optimized advertisement putting time period suggestion. In this way, the advertisement putting effect and efficiency are remarkably improved, and scientific decision support is provided for brands and advertisers.

Description

Technical Field

[0001] This application relates to the technical field of advertising management, and more specifically, to an urban advertising placement management system and method. Background Art

[0002] In modern urban advertising placement management, with the rapid development of the digital advertising market and the diversification of consumer behavior, how to efficiently and accurately place advertisements has become an important challenge for advertisers and brand managers. Traditionally, advertising placement strategies mainly rely on market research and historical data to determine the best placement time and location. However, this method has obvious limitations, especially when dealing with the rapidly changing market environment and competitive situation, it seems powerless.

[0003] First of all, traditional advertising placement management methods usually lack effective consideration of dynamic external environmental factors. For example, external variables such as temperature, humidity, and holidays will have an important impact on consumers' purchasing behavior, but these factors are often not fully incorporated into the advertising placement decision-making process. In addition, although advertising effectiveness indicators such as click-through rate and conversion rate can provide certain feedback information, they only reflect the attractiveness of the advertisement itself and ignore the influence of potential consumers by other external factors. This single-dimensional analysis method is difficult to comprehensively reflect the true effect of advertising placement, resulting in unreasonable allocation of advertising resources and reducing the return on advertising investment. Secondly, existing technologies also have deficiencies in dealing with the advertising placement intensity of competing products. As market competition becomes increasingly fierce, competitors may also increase their advertising placement efforts during the same period, which will undoubtedly affect the effect of the target advertisement. However, traditional advertising placement management systems often cannot timely capture this competitive intelligence, let alone effectively analyze it and adjust their own advertising strategies accordingly. Therefore, even with high-quality advertising content, in a fierce market competition environment, it may lose its competitive advantage due to the inability to accurately grasp market dynamics.

[0004] Therefore, an optimized urban advertising placement management solution is expected. Summary of the Invention

[0005] To solve the above technical problems, this application is proposed. Embodiments of this application provide an urban advertising placement management system and method, aiming to optimize advertising placement strategies through advanced machine learning technologies. To solve the problems in the prior art that there are obvious deficiencies in dealing with dynamic external environmental factors and the advertising placement intensity of competing products, resulting in inaccurate prediction of advertising effects and unreasonable allocation of resources.

[0006] According to one aspect of the present application, a city advertising placement management system is provided, including: a time queue for inputting advertising effect indicators within a predetermined time window and a time queue for external environmental variables, where the predetermined time window is 7 days; a time queue for inputting competitive intelligence within the predetermined time window; performing time series encoding on the time queue of advertising effect indicators, the time queue of external environmental variables, and the time queue of competitive intelligence based on a double-layer LSTM network to obtain an advertising effect indicator time-series periodic pattern feature encoding vector, an external environmental variable time-series periodic pattern feature encoding vector, and a competitive intelligence time-series periodic pattern feature encoding vector, wherein the first layer LSTM of the double-layer LSTM network is used to capture intra-day time-series periodic pattern features, and the second layer LSTM of the double-layer LSTM network is used to capture cross-day time-series periodic pattern features; inputting the advertising effect indicator time-series periodic pattern feature encoding vector, the external environmental variable time-series periodic pattern feature encoding vector, and the competitive intelligence time-series periodic pattern feature encoding vector into a dynamic decision module based on the Softmax function to obtain an advertising placement prediction result; generating a suggestion for the advertising placement time period based on the advertising placement prediction result.

[0007] In the above-mentioned city advertising placement management system, the advertising effect indicators include click-through rate and conversion rate, the external environmental variables include temperature, humidity, and holiday flag bits, and the competitive intelligence is the advertising placement intensity of competing products within 500 meters.

[0008] In the above-mentioned city advertising placement management system, the first layer LSTM contains 64 neurons, and the second layer LSTM contains 32 neurons.

[0009] In the above urban advertising placement management system, time series encoding is performed on the time queue of the advertising effect indicators, the time queue of the external environmental variables, and the time queue of the competitive intelligence based on a double-layer LSTM network to obtain an advertising effect indicator time-series periodic pattern feature encoding vector, an external environmental variable time-series periodic pattern feature encoding vector, and a competitive intelligence time-series periodic pattern feature encoding vector, including: inputting the time queue of the advertising effect indicators, the time queue of the external environmental variables, and the time queue of the competitive intelligence into the first-layer LSTM of the double-layer LSTM network to obtain a sequence distribution of the advertising effect indicator within-day periodic pattern feature encoding vector, a sequence distribution of the external environmental variable within-day periodic pattern feature encoding vector, and a sequence distribution of the competitive intelligence within-day periodic pattern feature encoding vector; respectively inputting the sequence distribution of the advertising effect indicator within-day periodic pattern feature encoding vector, the sequence distribution of the external environmental variable within-day periodic pattern feature encoding vector, and the sequence distribution of the competitive intelligence within-day periodic pattern feature encoding vector into the second-layer LSTM of the double-layer LSTM network to obtain the advertising effect indicator time-series periodic pattern feature encoding vector, the external environmental variable time-series periodic pattern feature encoding vector, and the competitive intelligence time-series periodic pattern feature encoding vector.

[0010] In the above urban advertising placement management system, inputting the advertising effect indicator time-series periodic pattern feature encoding vector, the external environmental variable time-series periodic pattern feature encoding vector, and the competitive intelligence time-series periodic pattern feature encoding vector into a dynamic decision module based on the Softmax function to obtain an advertising placement prediction result, including: inputting the advertising effect indicator time-series periodic pattern feature encoding vector, the external environmental variable time-series periodic pattern feature encoding vector, and the competitive intelligence time-series periodic pattern feature encoding vector into a triple attention network to obtain an enhanced advertising effect indicator time-series periodic pattern feature encoding vector; fusing the enhanced advertising effect indicator time-series periodic pattern feature encoding vector, the external environmental variable time-series periodic pattern feature encoding vector, and the competitive intelligence time-series periodic pattern feature encoding vector to obtain an advertising placement multi-source time-series fusion encoding vector; Inputting the advertising placement multi-source time-series fusion encoding vector into the dynamic decision module based on the Softmax function to obtain the advertising placement prediction result.

[0011] In the above urban advertising placement management system, inputting the time-series periodic pattern feature encoding vector of the advertising effect index, the time-series periodic pattern feature encoding vector of the external environmental variables, and the time-series periodic pattern feature encoding vector of the competitive intelligence into a triple attention network to obtain an enhanced time-series periodic pattern feature encoding vector of the advertising effect index includes: calculating a feature fine-grained association matrix of the time-series periodic pattern feature encoding vector of the external environmental variables and the time-series periodic pattern feature encoding vector of the competitive intelligence to obtain an external environment-competitive intelligence time-series collaborative attention matrix; performing feature activation based on the Sigmoid function on the external environment-competitive intelligence time-series collaborative attention matrix to obtain an external environment-competitive intelligence time-series collaborative attention mask matrix; mapping the time-series periodic pattern feature encoding vector of the advertising effect index to the external environment-competitive intelligence time-series collaborative attention mask matrix to obtain an attention-modulated time-series periodic pattern feature encoding vector of the advertising effect index; and inputting the attention-modulated time-series periodic pattern feature encoding vector of the advertising effect index and the time-series periodic pattern feature encoding vector of the advertising effect index into a residual unit to obtain the enhanced time-series periodic pattern feature encoding vector of the advertising effect index.

[0012] In the above urban advertising placement management system, calculating a feature fine-grained association matrix of the time-series periodic pattern feature encoding vector of the external environmental variables and the time-series periodic pattern feature encoding vector of the competitive intelligence to obtain an external environment-competitive intelligence time-series collaborative attention matrix includes: calculating the feature fine-grained association matrix of the time-series periodic pattern feature encoding vector of the external environmental variables and the time-series periodic pattern feature encoding vector of the competitive intelligence with the following formula to obtain the external environment-competitive intelligence time-series collaborative attention matrix; where the formula is: ; where represents the time-series periodic pattern feature encoding vector of the external environmental variables, represents the time-series periodic pattern feature encoding vector of the competitive intelligence, represents matrix multiplication, represents the transpose operation, is the feature length of the time-series periodic pattern feature encoding vector of the external environmental variables and the time-series periodic pattern feature encoding vector of the competitive intelligence, represents the external environment-competitive intelligence time-series collaborative attention matrix.

[0013] In the above urban advertising placement management system, calculating a feature fine-grained association matrix of the external environmental variable time-series periodic pattern feature encoding vector and the competitive intelligence time-series periodic pattern feature encoding vector to obtain an external environment-competitive intelligence time-series collaborative attention matrix includes: performing distribution stability probability on the external environmental variable time-series periodic pattern feature encoding vector and the competitive intelligence time-series periodic pattern feature encoding vector based on the interactive global mean to obtain an external environmental variable time-series periodic pattern feature stability probability vector and a competitive intelligence time-series periodic pattern feature stability probability vector; calculating an interaction step size and a generalized interaction coefficient based on the feature means of the external environmental variable time-series periodic pattern feature encoding vector and the competitive intelligence time-series periodic pattern feature encoding vector; performing truncation approximation based on historical values on the external environmental variable time-series periodic pattern feature stability probability vector and the competitive intelligence time-series periodic pattern feature stability probability vector based on the calculated interaction step size and generalized interaction coefficient to obtain an external environmental variable time-series periodic pattern feature truncation approximation vector and a competitive intelligence time-series periodic pattern feature truncation approximation vector; calculating a feature fine-grained association matrix of the external environmental variable time-series periodic pattern feature truncation approximation vector and the competitive intelligence time-series periodic pattern feature truncation approximation vector to obtain an external environment-competitive intelligence time-series collaborative attention matrix.

[0014] In the above urban advertising placement management system, the advertising placement prediction result is the probability value of each time period as the best placement opportunity.

[0015] According to another aspect of the present application, there is provided an urban advertising placement management system, including: a time data input module for inputting a time queue of advertising effect indicators and a time queue of external environmental variables within a predetermined time window, where the predetermined time window is 7 days; a competitive intelligence input module for inputting a time queue of competitive intelligence within the predetermined time window; a periodic pattern feature extraction module for performing time series encoding on the time queue of advertising effect indicators, the time queue of external environmental variables, and the time queue of competitive intelligence based on a double-layer LSTM network to obtain an advertising effect indicator time series periodic pattern feature encoding vector, an external environmental variable time series periodic pattern feature encoding vector, and a competitive intelligence time series periodic pattern feature encoding vector, wherein the first layer LSTM of the double-layer LSTM network is used to capture intra-day time series periodic pattern features, and the second layer LSTM of the double-layer LSTM network is used to capture cross-day time series periodic pattern features; an advertising placement prediction result determination module for inputting the advertising effect indicator time series periodic pattern feature encoding vector, the external environmental variable time series periodic pattern feature encoding vector, and the competitive intelligence time series periodic pattern feature encoding vector into a dynamic decision module based on the Softmax function to obtain an advertising placement prediction result; and an advertising placement suggestion generation module for generating suggestions for advertising placement time periods based on the advertising placement prediction result.

[0016] Compared with the prior art, the urban advertising placement management system and method provided by the present application obtain the time queue of advertising effect indicators, the time queue of external environmental variables, and the time queue of competitive intelligence within a predetermined time window, and then use a double-layer LSTM network to perform time series encoding on advertising effect indicators, external environmental variables (such as temperature, humidity, holiday flag bits), and competitive intelligence (such as the advertising placement intensity of competing products within 500 meters), and capture intra-day and cross-day periodic pattern features. The feature representation is enhanced through a triple attention network, and multi-source time series data is fused to generate accurate advertising placement prediction results. The dynamic decision module based on the Softmax function further gives optimized suggestions for advertising placement time periods. In this way, the effect and efficiency of advertising placement are significantly improved, providing scientific decision-making support for brands and advertisers. Brief Description of the Drawings

[0017] By describing the embodiments of the present application in more detail in conjunction with the drawings, the above and other objects, features, and advantages of the present application will become more obvious. The drawings are used to provide a further understanding of the embodiments of the present application, and constitute a part of the specification, and are used to explain the present application together with the embodiments of the present application, and do not constitute a limitation to the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0018] Figure 1The flowchart shows the process of the urban advertising placement management method according to an embodiment of the present application.

[0019] Figure 2 The flowchart shows the process of step S3 in the urban advertising placement management method according to an embodiment of the present application.

[0020] Figure 3 The flowchart shows the process of step S4 in the urban advertising placement management method according to an embodiment of the present application.

[0021] Figure 4 The flowchart shows the process of step S41 in the urban advertising placement management method according to an embodiment of the present application.

[0022] Figure 5 The structural diagram shows the structure of the urban advertising placement management system according to an embodiment of the present application. Detailed implementation manners

[0023] Next, exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. It should be understood that the present application is not limited by the exemplary embodiments described herein.

[0024] Figure 1 The flowchart shows the process of the urban advertising placement management method according to an embodiment of the present application. As Figure 1 shown, the urban advertising placement management method includes: S1, inputting the time queues of advertising effect indicators and external environment variables within a predetermined time window, where the predetermined time window is 7 days; S2, inputting the time queue of competitive intelligence within the predetermined time window; S3, performing time series encoding on the time queues of the advertising effect indicators, the external environment variables, and the competitive intelligence based on a double-layer LSTM network to obtain an advertising effect indicator time-series periodic pattern feature encoding vector, an external environment variable time-series periodic pattern feature encoding vector, and a competitive intelligence time-series periodic pattern feature encoding vector, where the first layer LSTM of the double-layer LSTM network is used to capture the within-day time-series periodic pattern features, and the second layer LSTM of the double-layer LSTM network is used to capture the cross-day time-series periodic pattern features; S4, inputting the advertising effect indicator time-series periodic pattern feature encoding vector, the external environment variable time-series periodic pattern feature encoding vector, and the competitive intelligence time-series periodic pattern feature encoding vector into a dynamic decision module based on the Softmax function to obtain an advertising placement prediction result; S5, generating a suggestion for the advertising placement time period based on the advertising placement prediction result.

[0025] Specifically, in step S1, a time queue of advertising effect metrics and a time queue of external environmental variables within a predetermined time window are input. The predetermined time window is 7 days. In one example, the advertising effect metrics include click-through rate and conversion rate, and the external environmental variables include temperature, humidity, and holiday flag. Specifically, first, the time queue of advertising effect metrics is considered. The advertising effect metrics referred to here usually include key performance indicators such as click-through rate (CTR) and conversion rate (CVR). The click-through rate (CTR) and conversion rate (CVR) are aggregated according to time periods to form a seven-day time series. In one embodiment, time series data is obtained on an hourly basis. Of course, the time period can also be selected according to the type of advertisement. In this way, the changing trend of the advertising effect over time can be clearly observed, such as whether there is a significant increase or decrease on weekends or specific holidays. For example, at the same time, the time queue of external environmental variables is also crucial. Such variables may cover various factors such as temperature, humidity, and holiday flag. Therefore, meteorological data such as temperature and humidity for each time period and information such as whether it is a holiday need to be recorded. These data are also organized according to a seven-day time window to form a time series of external environmental variables. This approach helps to identify how external conditions indirectly affect the performance of advertisements.

[0026] The reason for combining the above two types of time series data is that they can jointly reveal the complex dynamic relationships behind the advertising effect. The choice of using 7 days as a predetermined time window is not arbitrary. On the one hand, 7 days is long enough to capture intra-day and cross-day pattern changes; on the other hand, it is not too long to lose timeliness. For example, in a rapidly changing market environment, an overly long time window may lead to a lag effect, making decisions based on old data no longer applicable. Instead, a 7-day time window can reflect both short-term fluctuations and capture periodic trends.

[0027] Specifically, in step S2, a time queue of competitive intelligence within the predetermined time window is input. In one example, the competitive intelligence is the advertising placement intensity of competing products within 500 meters. It should be understood that market competition is a dynamic process, and the advertising activities of competitors may significantly affect the advertising effect. For example, if it is found that the number of competing product ad displays increases significantly in a day, then even if the advertisement content is attractive, the click-through rate and conversion rate may decrease due to the distraction of consumers' attention. By continuously monitoring these changes, the advertising placement strategy can be adjusted in a timely manner to avoid waste of resources.

[0028] In a specific implementation example, an effective data collection mechanism is established to obtain this competitive intelligence. This can be achieved through various means. For example, by using the data interface of a third-party advertising service provider to regularly scrape the display data of competitors' online advertisements; or by conducting on-site research to record the advertising arrangements of surrounding stores. Record this information every day and organize it into a seven-day time series. For example, in each time period of each day from Monday to Sunday, the total display times or distribution density of all competing product advertisements within 500 meters on that day are counted. In addition, organizing this competitive intelligence according to a seven-day time window helps to identify periodic patterns. For example, some competitors may increase their advertising efforts on weekends in an attempt to attract more people who shop for leisure on weekends. If this can be predicted in advance and the advertising placement plan is adjusted accordingly, an active position can be occupied in the fierce market competition.

[0029] Specifically, in step S3, based on a double-layer LSTM network, time series encoding is performed on the time queue of the advertising effect indicators, the time queue of the external environmental variables, and the time queue of the competitive intelligence to obtain an advertising effect indicator time series periodic pattern feature encoding vector, an external environmental variable time series periodic pattern feature encoding vector, and a competitive intelligence time series periodic pattern feature encoding vector. Among them, the first layer LSTM of the double-layer LSTM network is used to capture the intra-day time series periodic pattern features, and the second layer LSTM of the double-layer LSTM network is used to capture the cross-day time series periodic pattern features. In an example, the first layer LSTM contains 64 neurons, and the second layer LSTM contains 32 neurons. Specifically, the reason for using a double-layer LSTM network for time series encoding is that this method can effectively capture complex pattern features at different time scales. First of all, the changes in advertising effects, external environmental variables, and competitive intelligence often have multi-level dynamic characteristics. For example, the effect of an advertisement may fluctuate significantly within a day, but there are also cross-day trend changes. The first layer LSTM captures the intra-day pattern through a relatively large number of neurons (64), which can carefully reflect these short-term fluctuations; while the second layer LSTM captures the cross-day pattern through a relatively small number of neurons (32), which helps to identify long-term trends. In addition, the design of the double-layer LSTM network also takes into account the balance between computational efficiency and model complexity. The first layer LSTM captures the intra-day pattern through a relatively large number of neurons (64) to ensure that short-term fluctuations can be carefully reflected; while the second layer LSTM captures the cross-day pattern through a relatively small number of neurons (32), which not only ensures the generalization ability of the model but also avoids the risk of overfitting. This design enables high accuracy to be maintained while having good real-time performance and scalability.

[0030] In an example, such as Figure 2As shown, in step S3, time series encoding is performed on the time queue of the advertising effect indicators, the time queue of the external environmental variables, and the time queue of the competitive intelligence based on a double-layer LSTM network to obtain an advertising effect indicator time-series periodic pattern feature encoding vector, an external environmental variable time-series periodic pattern feature encoding vector, and a competitive intelligence time-series periodic pattern feature encoding vector, including: S31, inputting the time queue of the advertising effect indicators, the time queue of the external environmental variables, and the time queue of the competitive intelligence into the first layer LSTM of the double-layer LSTM network to obtain a sequence distribution of the advertising effect indicator within-day periodic pattern feature encoding vectors, a sequence distribution of the external environmental variable within-day periodic pattern feature encoding vectors, and a sequence distribution of the competitive intelligence within-day periodic pattern feature encoding vectors; S32, respectively inputting the sequence distribution of the advertising effect indicator within-day periodic pattern feature encoding vectors, the sequence distribution of the external environmental variable within-day periodic pattern feature encoding vectors, and the sequence distribution of the competitive intelligence within-day periodic pattern feature encoding vectors into the second layer LSTM of the double-layer LSTM network to obtain the advertising effect indicator time-series periodic pattern feature encoding vector, the external environmental variable time-series periodic pattern feature encoding vector, and the competitive intelligence time-series periodic pattern feature encoding vector.

[0031] Specifically, the time series of the advertising effect indicators, the time series of the external environmental variables, and the time series of the competitive intelligence are sequentially input into the first layer LSTM. The first layer LSTM processes these data day by day and extracts the within-day pattern features in each time series. For example, it may find that the click-through rate of certain advertisements is particularly high around 8 pm, or the conversion rate on weekends is significantly higher than on weekdays. These within-day pattern features are encoded into a series of vectors, forming a sequence distribution of the advertising effect indicator within-day periodic pattern feature encoding vectors, the external environmental variable within-day periodic pattern feature encoding vectors, and the competitive intelligence within-day periodic pattern feature encoding vectors. Then, the sequence distribution of these within-day periodic pattern feature encoding vectors is input into the second layer LSTM. The role of the second layer LSTM is to capture the cross-day pattern features, that is, to identify the trend changes between different dates within a week. For example, it may find that the advertising effect is better in the first half of a certain week, but the effect decreases in the second half due to the increased advertising investment of competitors. In this way, the second layer LSTM can generate more comprehensive advertising effect indicator time-series periodic pattern feature encoding vectors, external environmental variable time-series periodic pattern feature encoding vectors, and competitive intelligence time-series periodic pattern feature encoding vectors.

[0032] Specifically, in step S4, the advertising effect indicator time-series periodic pattern feature encoding vector, the external environmental variable time-series periodic pattern feature encoding vector, and the competitive intelligence time-series periodic pattern feature encoding vector are input into a dynamic decision-making module based on the Softmax function to obtain an advertising placement prediction result.

[0033] In one example, as Figure 3 shown, in step S4, inputting the time-series periodic pattern feature encoding vector of the advertisement effect index, the time-series periodic pattern feature encoding vector of the external environmental variable, and the time-series periodic pattern feature encoding vector of the competitive intelligence into a dynamic decision module based on the Softmax function to obtain an advertisement placement prediction result, including: S41, inputting the time-series periodic pattern feature encoding vector of the advertisement effect index, the time-series periodic pattern feature encoding vector of the external environmental variable, and the time-series periodic pattern feature encoding vector of the competitive intelligence into a triple attention network to obtain an enhanced time-series periodic pattern feature encoding vector of the advertisement effect index; S42, fusing the enhanced time-series periodic pattern feature encoding vector of the advertisement effect index, the time-series periodic pattern feature encoding vector of the external environmental variable, and the time-series periodic pattern feature encoding vector of the competitive intelligence to obtain a multi-source time-series fusion encoding vector for advertisement placement; S43, inputting the multi-source time-series fusion encoding vector for advertisement placement into the dynamic decision module based on the Softmax function to obtain the advertisement placement prediction result.

[0034] Specifically, in step S41, when predicting the advertisement placement effect, the influence mechanisms of external environmental factors and the advertisement placement intensity of competing products on the advertisement placement effect are different, and there is also an implicit pattern association between external environmental factors and the advertisement placement intensity of competing products. Therefore, in order to optimize the accuracy of advertisement placement prediction, the triple attention network is constructed. Specifically, the design of the triple attention network aims to enhance the time-series periodic pattern feature encoding vector of the advertisement effect index. The advertisement effect itself is a complex concept. It not only depends on the quality of the advertisement content but is also affected by the external environment and competitors. By introducing the attention mechanism, it can automatically identify which external factors have the greatest impact on the advertisement effect and accordingly adjust the advertisement placement strategy.

[0035] In one example, as Figure 4As shown, in step S41, inputting the time-series periodic pattern feature encoding vector of the advertisement effect index, the time-series periodic pattern feature encoding vector of the external environment variable, and the time-series periodic pattern feature encoding vector of the competitive intelligence into the triple attention network to obtain an enhanced time-series periodic pattern feature encoding vector of the advertisement effect index, including: S411, calculating the feature fine-grained association matrix of the time-series periodic pattern feature encoding vector of the external environment variable and the time-series periodic pattern feature encoding vector of the competitive intelligence to obtain an external environment-competitive intelligence time-series collaborative attention matrix; S412, performing feature activation based on the Sigmoid function on the external environment-competitive intelligence time-series collaborative attention matrix to obtain an external environment-competitive intelligence time-series collaborative attention mask matrix; S413, mapping the time-series periodic pattern feature encoding vector of the advertisement effect index to the external environment-competitive intelligence time-series collaborative attention mask matrix to obtain an attention-modulated time-series periodic pattern feature encoding vector of the advertisement effect index; S414, inputting the attention-modulated time-series periodic pattern feature encoding vector of the advertisement effect index and the time-series periodic pattern feature encoding vector of the advertisement effect index into a residual unit to obtain the enhanced time-series periodic pattern feature encoding vector of the advertisement effect index.

[0036] In one example, calculating the feature fine-grained association matrix of the time-series periodic pattern feature encoding vector of the external environment variable and the time-series periodic pattern feature encoding vector of the competitive intelligence to obtain an external environment-competitive intelligence time-series collaborative attention matrix includes: calculating the feature fine-grained association matrix of the time-series periodic pattern feature encoding vector of the external environment variable and the time-series periodic pattern feature encoding vector of the competitive intelligence with the following formula to obtain the external environment-competitive intelligence time-series collaborative attention matrix; where the formula is: ; where represents the time-series periodic pattern feature encoding vector of the external environment variable, represents the time-series periodic pattern feature encoding vector of the competitive intelligence, represents matrix multiplication, represents the transpose operation, is the feature length of the time-series periodic pattern feature encoding vector of the external environment variable and the time-series periodic pattern feature encoding vector of the competitive intelligence, represents the external environment-competitive intelligence time-series collaborative attention matrix.

[0037] Specifically, first, it is necessary to calculate the feature fine-grained correlation matrix between the temporal periodic pattern feature encoding vectors of the external environmental variables and the temporal periodic pattern feature encoding vectors of the competitive intelligence. Then, perform feature activation on the external environment-competitive intelligence temporal collaborative attention matrix based on the Sigmoid function to obtain the external environment-competitive intelligence temporal collaborative attention mask matrix. The role of the Sigmoid function is to map numerical values between 0 and 1, making certain features more prominent while suppressing other features. This can better capture the implicit relationship between external environmental variables and competitive intelligence. Map the temporal periodic pattern feature encoding vector of the advertising effect index to the above attention mask matrix to obtain the attention-modulated temporal periodic pattern feature encoding vector of the advertising effect index. The purpose of this step is to enable the advertising effect index to better reflect the influence of the external environment and competitive situation. For example, if the temperature is low and the competitor's advertising investment intensity is high on a certain day, then the advertising effect index on that day may be greatly affected, and attention modulation can help identify this influence. Input the attention-modulated temporal periodic pattern feature encoding vector of the advertising effect index and the original temporal periodic pattern feature encoding vector of the advertising effect index into the residual unit together. The role of the residual unit is to introduce new attention modulation information while retaining the original information, thereby obtaining an enhanced temporal periodic pattern feature encoding vector of the advertising effect index.

[0038] In a preferred example, calculating the feature fine-grained correlation matrix of the temporal periodic pattern feature encoding vector of the external environmental variables and the temporal periodic pattern feature encoding vector of the competitive intelligence to obtain the external environment-competitive intelligence temporal collaborative attention matrix includes: performing distribution stability probability on the temporal periodic pattern feature encoding vector of the external environmental variables and the temporal periodic pattern feature encoding vector of the competitive intelligence based on the interactive global mean to obtain the stable probability vectors of the temporal periodic pattern features of the external environmental variables and the stable probability vectors of the temporal periodic pattern features of the competitive intelligence; calculating the interaction step and the generalized interaction coefficient based on the feature means of the temporal periodic pattern feature encoding vector of the external environmental variables and the temporal periodic pattern feature encoding vector of the competitive intelligence; performing truncation approximation based on historical values on the stable probability vectors of the temporal periodic pattern features of the external environmental variables and the stable probability vectors of the temporal periodic pattern features of the competitive intelligence according to the calculated interaction step and generalized interaction coefficient to obtain the truncated approximation vectors of the temporal periodic pattern features of the external environmental variables and the truncated approximation vectors of the temporal periodic pattern features of the competitive intelligence; calculating the feature fine-grained correlation matrix of the truncated approximation vectors of the temporal periodic pattern features of the external environmental variables and the truncated approximation vectors of the temporal periodic pattern features of the competitive intelligence to obtain the external environment-competitive intelligence temporal collaborative attention matrix.

[0039] Specifically, first, perform distribution stability probability on the encoding vector of the time-series periodic pattern features of the external environmental variables and the encoding vector of the time-series periodic pattern features of the competitive intelligence, that is: ; where and are the feature means of the encoding vector of the time-series periodic pattern features of the external environmental variables and the encoding vector of the time-series periodic pattern features of the competitive intelligence respectively, represents vector addition, represents the Hadamard product, represents calculating the reciprocal of each eigenvalue in the vector, is the stable probability vector of the time-series periodic pattern features of the external environmental variables, is the stable probability vector of the time-series periodic pattern features of the competitive intelligence.

[0040] Then, obtain the interaction step size of the encoding vector of the time-series periodic pattern features of the external environmental variables and the encoding vector of the time-series periodic pattern features of the competitive intelligence through , and further calculate the generalized interaction coefficient: ; where represents rounding up, represents the interaction step size, represents factorial, is the generalized interaction coefficient for the encoding vector of the time-series periodic pattern features of the external environmental variables, is the generalized interaction coefficient for the encoding vector of the time-series periodic pattern features of the competitive intelligence; thus, infer the interaction long-tail distribution elastically based on historical interaction non-locality.

[0041] Finally, perform truncation approximation based on historical values in an asymptotic recurrence manner: ; where represents the truncation approximation vector of the time-series periodic pattern features of the external environmental variables, represents the truncation approximation vector of the time-series periodic pattern features of the competitive intelligence.

[0042] That is, through the encoding vector of the time-series periodic pattern features of the external environmental variables and the encoding vector of the time-series periodic pattern features of the competitive intelligence Use the stable global historical interaction as a boundary condition, and infer the non-local long-tail distribution globally through truncation approximation, so as to achieve explicit full-history discretized interaction and compensate for the possible problem of insufficient explicit inference of the global history in the fine-grained association of features.

[0043] Specifically, in step S42, fuse the enhanced advertising effect metric time-series periodic pattern feature encoding vector, the external environment variable time-series periodic pattern feature encoding vector, and the competitive intelligence time-series periodic pattern feature encoding vector to obtain a multi-source time-series fusion encoding vector for advertising placement. Specifically, perform operations such as weighted summation or concatenation on these three encoding vectors to form a multi-dimensional vector representation. This vector not only contains the features of the advertising effect itself but also integrates the information of the external environment and the competitive situation, thus providing a more comprehensive perspective. That is, the generation of the multi-source time-series fusion encoding vector is to integrate various types of data and provide a more comprehensive feature representation. There may be complex interaction relationships between different data sources, and it is difficult to comprehensively understand the influencing factors of the advertising effect by relying solely on a certain type of data. By fusing the time-series encoding vectors of advertising effect metrics, external environment variables, and competitive intelligence, a richer feature representation can be extracted, thereby generating a more accurate advertising placement prediction result.

[0044] Specifically, in step S43, the dynamic decision-making module based on the Softmax function is used to generate an advertising placement prediction result according to the fused encoding vector. The advantage of the Softmax function is that it can clearly present the probability distribution of multiple candidate solutions, facilitating subsequent decision-making. For example, when faced with multiple different advertising placement time period options, the Softmax function can give the probability value of each time period as the best placement time, helping advertisers select the optimal placement time. In addition, the dynamic decision-making module can also be dynamically adjusted according to real-time data and historical data to ensure that the advertising placement strategy can quickly respond to market changes. This flexibility enables it to have good real-time performance and scalability while maintaining high accuracy.

[0045] The specific steps for the dynamic decision-making module based on the Softmax function to generate an advertising placement prediction result according to the fused encoding vector are as follows: The Softmax function maps the multi-source time-series fusion encoding vector to the probability distribution of each possible advertising placement time period. The probability of each time period reflects the possibility of that time period being the best placement time. For example, if there are multiple different advertising placement time period options, the Softmax function will give the probability value of each time period as the best placement time.

[0046] Here, the design of the ternary attention network and the dynamic decision-making module based on the Softmax function also takes into account the balance between computational efficiency and model complexity. The ternary attention network can significantly improve the quality of feature representation while keeping the model complexity relatively low by introducing the attention mechanism; while the dynamic decision-making module based on the Softmax function provides a clear and intuitive decision-making basis through the probability distribution. This design enables better real-time performance and scalability while maintaining high accuracy.

[0047] Specifically, in step S5, based on the advertisement placement prediction result, suggestions for the advertisement placement time period are generated. It should be understood that the dynamic decision-making module based on the Softmax function will give the probability value of each time period as the best placement opportunity. Based on the advertisement placement prediction result, specific suggestions for the advertisement placement time period can be generated. Specifically, those time periods with higher probability values will be recommended as the priority placement time periods. In addition, dynamic adjustment can also be made according to market changes and real-time data. For example, if it is found that the probability value of a certain time period suddenly drops (possibly due to increased advertisement placement efforts by competitors or changes in weather conditions), the suggestions can be adjusted in a timely manner to avoid waste of resources. For example, if the probability value of a certain time period throughout the day drops from 55% to 45%, it may be recommended to reduce the advertisement placement frequency in this time period and instead increase the placement in other high-success-rate time periods.

[0048] This application also provides an urban advertisement placement management system for executing the above-mentioned urban advertisement placement management method. Figure 5 The figure shows a schematic structural diagram of the urban advertisement placement management system according to an embodiment of this application. As Figure 5As shown, the urban advertising placement management system 500 includes: a time data input module 510 for inputting a time queue of advertising effect indicators and a time queue of external environmental variables within a predetermined time window, where the predetermined time window is 7 days; a competitive intelligence input module 520 for inputting a time queue of competitive intelligence within the predetermined time window; a periodic pattern feature extraction module 530 for performing time series encoding on the time queue of advertising effect indicators, the time queue of external environmental variables, and the time queue of competitive intelligence based on a double-layer LSTM network to obtain an advertising effect indicator time-series periodic pattern feature encoding vector, an external environmental variable time-series periodic pattern feature encoding vector, and a competitive intelligence time-series periodic pattern feature encoding vector. Among them, the first layer LSTM of the double-layer LSTM network is used to capture intra-day time-series periodic pattern features, and the second layer LSTM of the double-layer LSTM network is used to capture cross-day time-series periodic pattern features; an advertising placement prediction result determination module 540 for inputting the advertising effect indicator time-series periodic pattern feature encoding vector, the external environmental variable time-series periodic pattern feature encoding vector, and the competitive intelligence time-series periodic pattern feature encoding vector into a dynamic decision module based on the Softmax function to obtain an advertising placement prediction result; an advertising placement recommendation generation module 550 for generating a recommendation for the advertising placement time period based on the advertising placement prediction result.

[0049] Here, those skilled in the art can understand that the specific operations of each module in the above urban advertising placement management system have been described in detail above with reference to Figures 1 to 4 the description of the urban advertising placement management method, and therefore, the repeated description thereof will be omitted.

[0050] The embodiment of the present application also provides a computer-readable storage medium in which computer program code is stored. When the computer program code runs on a computer, the computer is caused to execute the above-related method steps to implement a urban advertising placement management method provided in the above embodiment.

[0051] The embodiment of the present application also provides a computer program product. When the computer program product runs on a computer, the computer is caused to execute the above-related steps to implement a urban advertising placement management method provided in the above embodiment.

[0052] Among them, the system, computer-readable storage medium, or computer program product provided by the embodiment of the present application are all used to execute the corresponding method provided above. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding method provided above, and will not be elaborated here.

[0053] It should be noted that the above-mentioned sequence of embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments.

[0054] The processes depicted in the accompanying drawings do not necessarily require the particular order or sequential order shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous. Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. The key point of each embodiment is to illustrate the differences from other embodiments.

Claims

1. A city advertising management method, characterized in that: include: Input a time queue of advertising effect indicators and a time queue of external environmental variables within a predetermined time window, wherein the predetermined time window is 7 days; Input the time queue of competitive intelligence within the predetermined time window; perform time series encoding on the time queue of the advertising effectiveness indicator, the time queue of the external environment variable and the time queue of the competitive intelligence based on a double-layer LSTM network to obtain a coding vector of the advertising effectiveness indicator temporal periodic pattern feature, a coding vector of the external environment variable temporal periodic pattern feature and a coding vector of the competitive intelligence temporal periodic pattern feature, wherein the first layer LSTM of the double-layer LSTM network is used to capture the temporal periodic pattern features within the day, and the second layer LSTM of the double-layer LSTM network is used to capture the temporal periodic pattern features across days; The advertising effect indicator temporal periodic pattern feature coding vector, the external environment variable temporal periodic pattern feature coding vector and the competitive intelligence temporal periodic pattern feature coding vector are input into a dynamic decision module based on a Softmax function to obtain an advertising delivery prediction result; based on the advertising delivery prediction result, a recommendation for an advertising delivery time period is generated.

2. The city advertising management method according to claim 1, characterized in that: The advertising effectiveness indicators include click-through rate and conversion rate, the external environmental variables include temperature, humidity and holiday flags, and the competitive intelligence is the intensity of competitor advertising within 500 meters.

3. The city advertising management method according to claim 2, characterized in that: The first LSTM layer contains 64 neurons, and the second LSTM layer contains 32 neurons.

4. The city advertising management method according to claim 3 is characterized in that: Based on a double-layer LSTM network, the time queue of the advertising effect indicator, the time queue of the external environment variable and the time queue of the competitive intelligence are time series encoded to obtain advertising effect indicator time series periodic pattern feature encoding vectors, external environment variable time series periodic pattern feature encoding vectors and competitive intelligence time series periodic pattern feature encoding vectors, including: inputting the time queue of the advertising effect indicator, the time queue of the external environment variable and the time queue of the competitive intelligence into the first layer LSTM of the double-layer LSTM network to obtain the sequence distribution of the advertising effect indicator daily periodic pattern feature encoding vectors, the sequence distribution of the external environment variable daily periodic pattern feature encoding vectors and the sequence distribution of the competitive intelligence daily periodic pattern feature encoding vectors; inputting the sequence distribution of the advertising effect indicator daily periodic pattern feature encoding vectors, the sequence distribution of the external environment variable daily periodic pattern feature encoding vectors and the sequence distribution of the competitive intelligence daily periodic pattern feature encoding vectors into the second layer LSTM of the double-layer LSTM network respectively to obtain the advertising effect indicator time series periodic pattern feature encoding vectors, the external environment variable time series periodic pattern feature encoding vectors and the competitive intelligence time series periodic pattern feature encoding vectors.

5. The city advertising management method according to claim 4 is characterized in that: The advertising effect indicator temporal periodic pattern feature coding vector, the external environment variable temporal periodic pattern feature coding vector and the competitive intelligence temporal periodic pattern feature coding vector are input into a dynamic decision module based on a Softmax function to obtain an advertising placement prediction result, including: inputting the advertising effect indicator temporal periodic pattern feature coding vector, the external environment variable temporal periodic pattern feature coding vector and the competitive intelligence temporal periodic pattern feature coding vector into a ternary attention network to obtain an enhanced advertising effect indicator temporal periodic pattern feature coding vector; fusing the enhanced advertising effect indicator temporal periodic pattern feature coding vector, the external environment variable temporal periodic pattern feature coding vector and the competitive intelligence temporal periodic pattern feature coding vector to obtain an advertising placement multi-source temporal fusion coding vector; and inputting the advertising placement multi-source temporal fusion coding vector into the dynamic decision module based on a Softmax function to obtain the advertising placement prediction result.

6. The city advertising management method according to claim 5, characterized in that: The advertising effect indicator temporal periodic pattern feature coding vector, the external environment variable temporal periodic pattern feature coding vector and the competitive intelligence temporal periodic pattern feature coding vector are input into a ternary attention network to obtain an enhanced advertising effect indicator temporal periodic pattern feature coding vector, including: calculating the feature fine-grained association matrix of the external environment variable temporal periodic pattern feature coding vector and the competitive intelligence temporal periodic pattern feature coding vector to obtain an external environment-competitive intelligence temporal collaborative attention matrix; performing feature activation based on a Sigmoid function on the external environment-competitive intelligence temporal collaborative attention matrix to obtain an external environment-competitive intelligence temporal collaborative attention mask matrix; mapping the advertising effect indicator temporal periodic pattern feature coding vector to the external environment-competitive intelligence temporal collaborative attention mask matrix to obtain an attention-modulated advertising effect indicator temporal periodic pattern feature coding vector; and inputting the attention-modulated advertising effect indicator temporal periodic pattern feature coding vector and the advertising effect indicator temporal periodic pattern feature coding vector into a residual unit to obtain the enhanced advertising effect indicator temporal periodic pattern feature coding vector.

7. The city advertising management method according to claim 6, characterized in that: Calculating the characteristic fine-grained association matrix of the external environment variable temporal periodic pattern feature coding vector and the competitive intelligence temporal periodic pattern feature coding vector to obtain the external environment-competitive intelligence temporal collaborative attention matrix, including: calculating the characteristic fine-grained association matrix of the external environment variable temporal periodic pattern feature coding vector and the competitive intelligence temporal periodic pattern feature coding vector using the following formula to obtain the external environment-competitive intelligence temporal collaborative attention matrix; wherein the formula is: ;in, Represents the characteristic encoding vector of the time series periodic pattern of the external environmental variable, represents the competitive intelligence temporal periodic pattern feature encoding vector, represents matrix multiplication, represents the transpose operation, is the characteristic length of the characteristic encoding vector of the time series periodic pattern of the external environment variable and the characteristic encoding vector of the time series periodic pattern of the competitive intelligence, Represents the external environment-competitive intelligence temporal collaborative attention matrix.

8. The city advertising management method according to claim 7, characterized in that: The method comprises: calculating the characteristic fine-grained association matrix of the time-series periodic pattern feature coding vector of the external environment variable and the time-series periodic pattern feature coding vector of the competitive intelligence to obtain the external environment-competitive intelligence time-series collaborative attention matrix, including: performing distribution stabilization probabilization on the time-series periodic pattern feature coding vector of the external environment variable and the time-series periodic pattern feature coding vector of the competitive intelligence based on the interactive global mean to obtain the external environment variable time-series periodic pattern feature stable probabilization vector and the competitive intelligence time-series periodic pattern feature stable probabilization vector; The characteristic mean of the code vector is calculated, and the interaction step and the generalized interaction coefficient are calculated; based on the calculated interaction step and the generalized interaction coefficient, the stable probabilistic vector of the characteristic periodic pattern of the external environment variable time series and the stable probabilistic vector of the characteristic periodic pattern of the competitive intelligence time series are truncated and approximated based on historical values ​​to obtain the truncated approximate vector of the characteristic periodic pattern of the external environment variable time series and the truncated approximate vector of the characteristic periodic pattern of the competitive intelligence time series; the characteristic fine-grained association matrix of the truncated approximate vector of the characteristic periodic pattern of the external environment variable time series and the truncated approximate vector of the characteristic periodic pattern of the competitive intelligence time series is calculated to obtain the external environment-competitive intelligence time series collaborative attention matrix.

9. The city advertising management method according to claim 8, characterized in that: The advertisement delivery prediction result is a probability value of each time period being the best delivery time.

10. A city advertisement delivery management system, used to execute the city advertisement delivery management method according to claims 1-9, characterized in that: include: A time data input module, used to input a time queue of advertising effectiveness indicators and a time queue of external environmental variables within a predetermined time window, wherein the predetermined time window is 7 days; a competitive intelligence input module, used to input a time queue of competitive intelligence within the predetermined time window; a periodic pattern feature extraction module, used to perform time series encoding on the time queue of advertising effectiveness indicators, the time queue of external environmental variables and the time queue of competitive intelligence based on a double-layer LSTM network to obtain an advertising effectiveness indicator time series periodic pattern feature encoding vector, an external environmental variable time series periodic pattern feature encoding vector and a competitive intelligence time series periodic pattern feature encoding vector, wherein the first LSTM layer of the double-layer LSTM network is used to capture intra-day time series periodic pattern features, and the second LSTM layer of the double-layer LSTM network is used to capture cross-day time series periodic pattern features; The advertising delivery prediction result determination module is used to input the advertising effect indicator temporal periodic pattern feature coding vector, the external environment variable temporal periodic pattern feature coding vector and the competitive intelligence temporal periodic pattern feature coding vector into a dynamic decision module based on the Softmax function to obtain the advertising delivery prediction result; the advertising delivery suggestion generation module is used to generate suggestions for the advertising delivery time period based on the advertising delivery prediction result.

Images