Method and system for automatically adjusting advertisement putting parameters based on advertisement effect

Through deep learning technology, the automatic adjustment of advertising delivery parameters is solved, and the complexity and inefficiency of online advertising delivery parameters are achieved, and the automated optimization and long-term stability of advertising effects are achieved, thereby reducing labor costs.

CN118967229APending Publication Date: 2024-11-15SHANGHAI GALLEON INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410997560.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-24
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

The online advertising delivery parameters are complex, relying on manual experience and trial and error, inefficient, and frequent changes in the market environment, resulting in parameter combination failure and frequent adjustments, which increases work complexity and difficulty.

Method used

Using a deep learning-based method, we use data preprocessing and training to determine the model, automatically adjust the advertising delivery parameters, use optimization strategies to generate parameter combinations, and iterate the training of the model to optimize the advertising effect.

Benefits of technology

It realizes automatic optimization of advertising effects, reduces the cost of manual parameter adjustment, improves parameter selection efficiency, adapts to market changes, maintains long-term and stable optimization effects, reduces labor costs, and improves advertising delivery results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118967229A_ABST
    Figure CN118967229A_ABST
Patent Text Reader

Abstract

The invention discloses a method and a system for automatically adjusting advertisement putting parameters based on an advertisement effect, which solve the blindness problem of manual experience parameter adjustment, reduce the manual parameter adjustment cost and quickly optimize the advertisement putting effect. According to the technical scheme, data preprocessing comprises data cleaning, data conversion and advertisement effect index and feature vector construction; training a judgment model based on a deep learning network, inputting feature vectors of different parameter combinations into the model, and outputting predicted values of new advertisement parameter combinations and advertisement effect indexes; using an optimization strategy to select a parameter combination corresponding to the highest value in the predicted values; using the parameter combination corresponding to the highest value in the advertisement effect prediction values to actually put the advertisement, and collecting the actual value of the advertisement effect index of the put advertisement; and inputting the actual value of the advertisement effect index into the judgment model, retraining the judgment model in an iteration mode, and storing the optimal parameter combination of the advertisement effect index after iteration training.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of Internet advertising delivery, and in particular to a method and system for automatically adjusting advertising delivery parameters based on advertising effects. Background Art

[0002] With the rapid advancement of Internet technology, online advertising has gradually become a core part of corporate marketing strategies and an important channel for corporate publicity and customer acquisition. However, this field is also accompanied by unprecedented challenges, especially the complexity of setting advertising parameters, which is directly related to the final effectiveness of advertising.

[0003] The success of online advertising is deeply affected by a combination of multiple parameters, including but not limited to the selection of delivery channels, the formulation of bidding strategies, precise crowd targeting, attractive copywriting design, visually appealing material production, and appropriate delivery time. The interactions between these parameters are intricate, and any slight adjustment may have a significant impact on the advertising effect.

[0004] Traditionally, the adjustment of advertising parameters relies on the experience and intuition of advertising personnel, who try to find the best parameter combination through trial and error. This method is not only time-consuming and labor-intensive, but also costly and inefficient. What is more difficult is that the constant changes in market environment, user preferences and competition situation may make the originally effective parameter combination quickly invalid, requiring frequent manual intervention for adjustment, which further increases the complexity and difficulty of the work. Summary of the invention

[0005] A brief summary of one or more aspects is given below to provide a basic understanding of these aspects. This summary is not an exhaustive overview of all conceived aspects, and is neither intended to identify the key or critical elements of all aspects nor to define the scope of any or all aspects. Its only purpose is to give some concepts of one or more aspects in a simplified form as a prelude to a more detailed description that will be given later.

[0006] The purpose of the present invention is to solve the above-mentioned problems and provide a method and system for automatically adjusting advertising delivery parameters based on advertising effects, which solves the blindness problem of manual experience-based parameter adjustment and enables the parameter combination to quickly converge to an optimized state, thereby reducing the cost of manual parameter adjustment and quickly optimizing the advertising delivery effect.

[0007] The technical solution of the present invention is: the present invention discloses a method for automatically adjusting advertisement delivery parameters based on advertisement effect, the method comprising:

[0008] Step 1: Data preprocessing, including data cleaning, data conversion, and constructing feature vectors for advertising effectiveness indicators and subsequent training models;

[0009] Step 2: training a decision model based on a deep learning network, wherein the input of the decision model is a feature vector of different parameter combinations, and the output of the decision model is a new advertising parameter combination and a predicted value of an advertising effect indicator;

[0010] Step 3: Use the optimization strategy to obtain the predicted values ​​of the advertising effect indicators corresponding to each parameter combination in the optimization strategy based on the judgment model, and select the parameter combination corresponding to the highest value;

[0011] Step 4: Use the parameter combination corresponding to the highest value in the advertising effect prediction value to actually place the advertisement, and collect the actual value of the advertising effect index of the placed advertisement;

[0012] Step 5: Input the actual value of the advertising effect index obtained in step 4 into the judgment model, retrain the judgment model in an iterative manner, and save the best parameter combination of the advertising effect index after iterative training.

[0013] According to an embodiment of the method for automatically adjusting advertising delivery parameters based on advertising effects of the present invention, data cleaning includes removing duplicate records and outliers, and processing missing values; data conversion includes converting categorical variables into numerical features and standardizing them so that the feature values ​​are on the same scale; advertising effect indicators include click-through rate, conversion rate, site rate, visit rate, or return on investment; construction of feature vectors includes first constructing a single feature and then constructing interaction terms between single features; feature vectors include: advertising copy, advertising platform, advertising material, delivery time, target population, advertising expenditure, display volume, click volume, visit volume, conversion volume and time series data.

[0014] According to an embodiment of the method for automatically adjusting advertisement delivery parameters based on advertisement effects of the present invention, in step 2, the deep learning network includes a long short-term memory network, a convolutional neural network, or a multi-layer perceptron.

[0015] According to an embodiment of the method for automatically adjusting advertisement delivery parameters based on advertisement effects of the present invention, in step 3, the optimization strategy includes a random strategy or a genetic strategy.

[0016] The present invention also discloses a system for automatically adjusting advertisement delivery parameters based on advertisement effects, the system comprising:

[0017] Data preprocessing module, used for data preprocessing, including data cleaning, data conversion, and constructing feature vectors for advertising effect indicators and subsequent training models;

[0018] A model training module is used to train a decision model based on a deep learning network, wherein the input of the decision model is a feature vector of different parameter combinations, and the output of the decision model is a new advertising parameter combination and a predicted value of an advertising effect indicator;

[0019] The optimization module uses the optimization strategy to obtain the predicted values ​​of the advertising effect indicators corresponding to each parameter combination in the optimization strategy based on the judgment model, and selects the parameter combination corresponding to the highest value;

[0020] The actual delivery module uses the parameter combination corresponding to the highest value in the advertising effect prediction value to actually deliver the advertisement and collect the actual value of the advertising effect index of the delivered advertisement;

[0021] The iterative training module inputs the actual values ​​of the advertising effect indicators obtained by the actual delivery module into the judgment model, retrains the judgment model in an iterative manner, and saves the best parameter combination of the advertising effect indicators after iterative training.

[0022] According to an embodiment of the system for automatically adjusting advertising delivery parameters based on advertising effectiveness of the present invention, in the data preprocessing module, data cleaning includes removing duplicate records and outliers, and processing missing values; data conversion includes converting categorical variables into numerical features and standardizing them so that the feature values ​​are on the same scale; advertising effectiveness indicators include click-through rate, conversion rate, site rate, visit rate, or return on investment; construction of feature vectors includes first constructing a single feature and then constructing interaction terms between single features; feature vectors include: advertising copy, advertising platform, advertising material, delivery time, target population, advertising expenditure, display volume, click volume, visit volume, conversion volume and time series data.

[0023] According to an embodiment of the system for automatically adjusting advertisement delivery parameters based on advertisement effects of the present invention, in the model training module, the deep learning network includes a long short-term memory network, a convolutional neural network, or a multi-layer perceptron.

[0024] According to an embodiment of the system for automatically adjusting advertisement delivery parameters based on advertisement effects of the present invention, in the optimization module, the optimization strategy includes a random strategy or a genetic strategy.

[0025] The present invention also discloses a computer-readable storage medium storing program instructions executable by a processor to implement the steps of the method for automatically adjusting advertisement delivery parameters based on advertisement effects as described above.

[0026] The present invention also discloses a computer device, which includes a memory, a processor and program instructions stored in the memory and executable by the processor, wherein the processor executes the program instructions to implement the steps of the method for automatically adjusting advertisement delivery parameters based on advertisement effects as described above.

[0027] Compared with the prior art, the present invention has the following beneficial effects: the present invention utilizes a machine learning algorithm to train a model, realizes automatic determination of advertising effects, and uses this to automatically adjust and optimize parameters. Specifically, the present invention collects advertising data, analyzes advertising effects under different parameter combinations, automatically generates new parameter combinations based on the advertising effects, re-releases advertising to verify the effects, and finally obtains the parameter combination with the best advertising effect. Compared with the prior art. The present invention helps to reduce advertising costs, improve the efficiency of parameter selection, and maintain long-term stable optimization effects.

[0028] In detail, the present invention can have the following advantages:

[0029] 1. Improved advertising effectiveness:

[0030] Through machine learning algorithm modeling, we can more accurately evaluate the advertising effects of different parameter combinations and find the optimal parameters, thereby improving advertising performance indicators such as click-through rate and conversion rate.

[0031] 2. Reduced manual parameter adjustment costs:

[0032] This method realizes the automation of parameter adjustment, does not require manual experience and repeated trial and error, and greatly reduces labor costs.

[0033] 3. Quickly respond to parameter adjustment needs:

[0034] Over time, this method can quickly retrain the model and automatically adjust parameters to adapt to parameter adjustment requirements brought about by changes in user preferences, etc.

[0035] 4. Parameter adjustment results can be explained:

[0036] By analyzing the impact of key parameters on advertising effectiveness, decision makers can understand the reasons for parameter adjustments and improve the transparency of the system.

[0037] 5. Continuously and stably optimize parameters:

[0038] This method implements closed-loop feedback and can continuously optimize parameters instead of a one-time parameter adjustment, thus ensuring long-term optimization effects.

[0039] 6. Reduce labor costs and improve advertising effects:

[0040] Overall, the biggest benefit of this method is that it reduces labor costs, improves advertising effectiveness, and enables companies to achieve a better input-output ratio. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] The above features and advantages of the present invention can be better understood after reading the detailed description of the embodiments of the present disclosure in conjunction with the following drawings. In the drawings, the components are not necessarily drawn to scale, and components with similar related properties or features may have the same or similar reference numerals.

[0042] Figure 1 A flow chart of an embodiment of a method for automatically adjusting advertisement delivery parameters based on advertisement effects of the present invention is shown.

[0043] Figure 2 A schematic diagram showing an embodiment of a system for automatically adjusting advertisement delivery parameters based on advertisement effects of the present invention.

[0044] Figure 3 A schematic diagram of the application process of connecting the system of the present invention to an e-commerce platform is shown. DETAILED DESCRIPTION

[0045] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. Note that the aspects described below in conjunction with the accompanying drawings and specific embodiments are only exemplary and should not be construed as limiting the scope of protection of the present invention in any way.

[0046] Figure 1 The flow chart of an embodiment of the method for automatically adjusting advertisement delivery parameters based on advertisement effect of the present invention is shown. Figure 1 , the implementation steps of the method of this embodiment are described in detail as follows.

[0047] Step 1: Data preprocessing.

[0048] In the process of advertising, raw data often has inconsistent and noisy characteristics, so the raw data must be preprocessed in a standardized manner to facilitate subsequent model training and prediction. Data preprocessing operations include: data cleaning, data conversion, and constructing feature vectors for advertising effect indicators (such as conversion rate, etc.) and subsequent training models.

[0049] Extract raw data from the advertising platform's MKTAPI interface, monitoring log files, and CRM systems, including advertising parameters, impressions, clicks, visits, conversions, etc. The MKTAPI interface (Marketing API) is a set of application programming interfaces (APIs) designed for marketers, advertisers, and developers to provide an efficient and automated way to manage and optimize advertising campaigns. Through the MKTAPI interface, users can access, query, modify, and report on various raw data about advertising campaigns. The CRM system (Customer Relationship Management System) is a software system used to manage the relationship between an enterprise and its customers. It not only records detailed customer information, but also integrates multiple business processes such as sales, marketing, and customer service.

[0050] An example of raw data is as follows: {Ad: A, "Distribution channel": Big Data Engine, "Bidding setting": Automatic bidding, "Targeted audience": Audience package X, "Copywriting": How to make summer cooler, "Material": Picture A, "Distribution time": 08:00-22:00, "Cost": 50, "Impressions": 1000, "Clicks": 50, "Visits": 40, "Conversions" 5, "Revenue": 2000}.

[0051] First, the original data is cleaned, including: removing duplicate records and outliers, and processing missing values. For duplicate records and outliers, for example, remove data with exactly the same advertising parameters and advertising effect data, and remove data entries with zero impressions but a conversion volume greater than 0. For missing values, for example, if the advertising parameters are consistent, but the number of visits to one of the ads is 0, and the other effect data is normal, the visit data can be filled by using methods such as mean filling or interpolation.

[0052] Then, the cleaned data is transformed, that is, the categorical variables are converted into numerical features and standardized so that the feature values ​​are on the same scale. For example, the text, material, time period, and crowd of the advertisement are encoded using One-Hot Encoding or Label Encoding. For the standardization of numerical features, for example, z-score standardization or Min-Max normalization is used to standardize the values.

[0053] Finally, based on the converted standardized numerical features, the advertising effectiveness indicators are defined and the feature vectors for the subsequent training model are constructed.

[0054] The defined advertising effectiveness indicators include: click-through rate (CTR), conversion rate (CVR), cost per unique visitor (CPUV), view-through rate (VTR), return on investment (ROI), etc.

[0055] The definition and calculation method of the above advertising effect indicators are as follows:

[0056] 1. Click-through rate (CTR):

[0057] Definition: The ratio of clicks to impressions, reflecting how often an ad is clicked by users;

[0058] Calculation formula: CTR = clicks / impressions;

[0059] Example: If the number of clicks is 50 and the number of impressions is 1000, then CTR = 50 / 1000 = 0.05;

[0060] 2. Conversion rate (CVR):

[0061] Definition: The ratio of conversions to clicks, which reflects the frequency of completing a specific action (such as purchase, registration, etc.) after clicking on an ad.

[0062] Calculation formula: CVR = conversion volume / click volume;

[0063] Example: If the conversion volume is 5 and the click volume is 50, then CVR = 5 / 50 = 0.1;

[0064] 3. Arrival rate (CPUV):

[0065] Definition: The ratio of the number of visits to the site to the number of clicks, reflecting the actual effect of traffic flow after clicking on the ad;

[0066] Calculation formula: CPUV = number of arrivals / number of clicks;

[0067] Example: If the number of arrivals is 5 and the number of clicks is 10, then CPUV = 5 / 10 = 0.5;

[0068] 4. Video Tracking Rate (VTR):

[0069] Definition: The ratio of visits to clicks, reflecting the frequency of visits to the target page after clicking on the ad;

[0070] Calculation formula: VTR = visits / clicks;

[0071] Example: If the number of visits is 40 and the number of clicks is 50, then VTR = 40 / 50 = 0.8;

[0072] 5. Return on Investment (ROI):

[0073] Definition: The ratio of revenue to advertising expenditure, reflecting the economic benefits of advertising;

[0074] Calculation formula: ROI = (revenue - cost) / cost;

[0075] Example: If the revenue is 2,000 yuan and the advertising cost is 50 yuan, then ROI = (2,000-50) / 1,000 = 1.5.

[0076] First, build a single feature, such as click-through rate (CTR), conversion rate (CVR), and CPUV.

[0077] Then construct interaction terms between features, such as the combined features of different delivery channels and time periods.

[0078] The feature fields of the example are as follows:

[0079] {Ad: A, "Distribution channel": Big Data Engine, "Bidding settings": Automatic bidding, "Crowd targeting": Crowd package X, "Copy": How to make summer cooler, "Material": Picture A, "Time": 08:00-09:00, "Position": Position A, "Cost": 50, "Displays": 1000, "Clicks": 50, "Station visits": 40, "Conversions": 5, "CTR": 0.05, "CVR": 0.125, "CPUV": 0.8};

[0080] {Ad: A, "Distribution channel": Big Data Engine, "Bidding setting": automatic bidding, "Targeted audience": audience package X, "Copy": How to make summer cooler, "Material": Picture A, "Time": 11:00-12:00, "Position": Position B, "Cost": 50, "Displays": 500, "Clicks": 30, "Visits": 15, "Conversions": 1, "CTR": 0.06, "CVR": 0.03, "CPUV": 0.5}.

[0081] Step 2: Train the judgment model.

[0082] A machine learning algorithm (such as LSTM long short-term memory network) is used to train the judgment model. The input of the judgment model is a feature vector of different parameter combinations, including advertising copy, advertising platform, advertising material, delivery time, target population, advertising expenditure, display volume, click volume, visit volume, conversion volume and time series data, etc. The output of the judgment model is a new advertising parameter combination and the predicted value of advertising effect indicators, such as click-through rate (CTR), conversion rate (CVR), CPUV, return on investment (ROI), etc.

[0083] The judgment model is essentially a deep learning model, and the best one is the long short-term memory network (LSTM) model. The long short-term memory network (LSTM) is a special recurrent neural network (RNN). LSTM can effectively capture the long-term and short-term dependencies in the data through its unique gating mechanism. In the prediction of advertising effect, LSTM can use the historical data of advertising delivery to predict the advertising effect under different parameter combinations in the future.

[0084] The following takes the LSTM model as an example to illustrate.

[0085] First, build a decision model. Build an LSTM model based on the data preprocessed in step 1, including the input layer, LSTM layer, and output layer of the model. A multi-layer LSTM structure is used to capture more complex time dependencies.

[0086] Then, the model is trained. The feature vector preprocessed in step 1 is input into the LSTM model, and the model is trained using historical data. The model is trained using an appropriate loss function (such as mean square error) and an optimization algorithm (such as the Adam optimizer). Hyperparameter tuning is performed to obtain optimal model performance.

[0087] Next, the model is evaluated. The prediction performance of the model is evaluated using the validation set data, with indicators including mean square error (MSE), root mean square error (RMSE), etc. The model structure and hyperparameters are adjusted based on the model evaluation results.

[0088] Finally, perform model prediction. Input the new parameter combination into the trained LSTM model, and output the new advertising parameter combination. Use the predicted advertising delivery results to perform parameter optimization and advertising delivery decisions.

[0089] The following is a simplified Python example code showing how to use the LSTM model to predict advertising effectiveness:

[0090]

[0091]

[0092] In addition to the above-mentioned LSTM neural network, many other deep learning models can also be used for advertising effect prediction, such as convolutional neural network (CNN) and multi-layer perceptron (MLP).

[0093] Take the CNN model as an example.

[0094] First, build a CNN model.

[0095] Then, the convolutional layer and pooling layer in the CNN model are constructed to extract local features in the time series data. A fully connected layer is added to integrate the features and output the prediction results.

[0096] After that, the CNN model training is the same as the LSTM model training process, which will not be repeated here.

[0097] Finally, the evaluation and prediction of the CNN model are the same as those of the LSTM model, so we will not go into details here.

[0098] The following is a simplified Python example code showing how to use the CNN model to predict advertising effectiveness:

[0099]

[0100]

[0101] Taking the multi-layer perceptron as an example, the multi-layer perceptron (MLP) is the most basic neural network structure, which is suitable for processing fixed-size input data. MLP performs feature extraction and nonlinear transformation through multiple hidden layers, and can capture complex relationships in advertising data.

[0102] First, build the MLP model.

[0103] Then, multiple fully connected layers in the MLP model are constructed to perform feature extraction and nonlinear transformation, and the output layer of the MLP model is constructed to predict advertising effect indicators.

[0104] After that, the MLP model training is carried out. The training process is the same as the LSTM model training process, which will not be repeated here.

[0105] Finally, the MLP model evaluation and prediction process is the same as the LSTM model evaluation and prediction process, which will not be repeated here.

[0106] The following is a simplified Python example code showing how to use the MLP model for advertising effect prediction:

[0107]

[0108]

[0109]

[0110] Step 3: Generate new parameter combinations, including: using an optimization strategy (including a random strategy or a genetic strategy), obtaining the predicted values ​​of the advertising effect indicators corresponding to each parameter combination in the optimization strategy based on the judgment model, and selecting the parameter combination corresponding to the highest value.

[0111] The implementation of the random strategy is as follows:

[0112] First, generate random parameter combinations. Define parameter space, such as ad copy, ad creative, delivery time, bid range, crowd targeting conditions, etc. Then randomly generate multiple parameter combinations.

[0113] Then, evaluate the effect of the parameter combination. Input the generated parameter combination into the trained prediction model to obtain the advertising effect index. Then select the parameter combination with the best effect.

[0114] Finally, save the optimal parameter combination. Save the parameter combination with the best effect for actual advertising.

[0115] The sample code for the random strategy is as follows.

[0116]

[0117]

[0118]

[0119] The implementation of the genetic strategy is as follows.

[0120] First, initialize the population. The initial population is randomly generated, and each individual represents a parameter combination.

[0121] Then, select the better individual. Select the better individual according to the individual's fitness (advertising effect).

[0122] After that, a crossover operation is performed to generate new individuals by exchanging some genes of the parent individuals.

[0123] Then, a mutation operation is performed to randomly change some genes of individuals to increase diversity.

[0124] Next, iterative optimization is performed, and selection, crossover, and mutation are repeated until the preset number of iterations is reached or the stopping condition is met.

[0125] Finally, save the optimal parameter combination. Save the parameter combination that works best during the iteration process.

[0126] The sample code for the genetic strategy is as follows.

[0127]

[0128]

[0129]

[0130] Step 4: Use the parameter combination corresponding to the highest value in the advertising effect prediction value to actually deliver the advertisement, and collect the actual value of the advertising effect indicator of the delivered advertisement.

[0131] Step 5: Input the actual value of the advertising effect index obtained in step 4 into the judgment model, retrain the judgment model in an iterative manner, and save the best parameter combination of the advertising effect index after iterative training.

[0132] Figure 2 The principle of an embodiment of the system for automatically adjusting advertisement delivery parameters based on advertisement effect of the present invention is shown. Figure 2 The system of this embodiment includes the following modules: data preprocessing module, model training module, optimization module, actual delivery module, and iterative training module.

[0133] The data preprocessing module is used to preprocess the raw data related to advertising, including data cleaning, data conversion, and constructing feature vectors of advertising effectiveness indicators (such as conversion rate, etc.) and subsequent training models.

[0134] For the technical content of the data preprocessing module, see Figure 1 The data preprocessing step 1 shown is not repeated here.

[0135] The model training module is used to use a machine learning algorithm to train a decision model, wherein the input of the decision model is a feature vector of different parameter combinations, and the output of the decision model is a new advertising parameter combination and a predicted value of an advertising effect indicator.

[0136] For technical content of the model training module, see Figure 1 Step 2 of the training judgment model shown is not repeated here.

[0137] The optimization module is used to generate new parameter combinations, including using optimization strategies (including random strategies or genetic strategies), obtaining predicted values ​​of advertising effectiveness indicators corresponding to each parameter combination in the optimization strategy based on the judgment model, and selecting the parameter combination corresponding to the highest value.

[0138] For technical content of the optimization module, see Figure 1 Step 3 of generating a new parameter combination is shown and will not be described in detail here.

[0139] The actual delivery module is used to actually deliver advertisements using the parameter combination corresponding to the highest value in the advertisement effect prediction value, and collect the actual values ​​of the advertisement effect indicators of the delivered advertisements.

[0140] For technical content of the actual delivery module, see Figure 1 The step 4 of "using the parameter combination corresponding to the highest value in the advertising effect prediction value to actually deliver the advertisement and collect the actual value of the advertising effect indicator of the delivered advertisement" is not repeated here.

[0141] The iterative training module is used to input the actual value of the advertising effect index into the judgment model, retrain the judgment model in an iterative manner, and save the best parameter combination of the advertising effect index after the iterative training.

[0142] For technical content of the iterative training module, see Figure 1 Step 5 of "inputting the actual value of the advertising effect indicator obtained in step 4 into the judgment model, retraining the judgment model in an iterative manner, and saving the best parameter combination of the advertising effect indicator after iterative training" is not repeated here.

[0143] Figure 3 The application process of connecting the system of the present invention to the e-commerce platform is shown. Figure 3 , the application process is as follows.

[0144] The first step is platform selection:

[0145] It is preferred to implement it on mainstream e-commerce advertising platforms, such as Tencent Advertising, Bytedance, Baidu Marketing, etc. These platforms have rich advertising data and better implementation effects.

[0146] The second step is algorithm selection:

[0147] The algorithm is mainly based on deep learning. You can try using the CNN model to extract features and the LSTM model to model time series to ensure prediction results.

[0148] The third step is indicator selection:

[0149] Taking conversion rate as the main evaluation indicator, taking into account indicators such as display volume, click volume, visit volume, ROI, etc., the advertising effect is comprehensively evaluated.

[0150] Step 4: Parameter selection:

[0151] It mainly optimizes seven core parameters: delivery channels, bidding, audience targeting, copywriting, materials, time periods, and placements. Other parameters can be expanded according to business needs.

[0152] Step 5: Data scale:

[0153] Prepare at least 100,000 pieces of historical advertising data with different parameter combinations to ensure the effectiveness of algorithm training.

[0154] Step 6, iteration speed:

[0155] Retrain the model every 48 hours to achieve rapid iterative optimization.

[0156] Step 7: Business docking:

[0157] Connect with advertising delivery systems and data platforms to establish automated data and model flows.

[0158] Step 8: Result analysis:

[0159] Analyze the impact of key parameters on results, track the performance data brought by new advertising parameter combinations and add them to the model training data to continuously optimize advertising parameters.

[0160] Step 9: Continuous Optimization:

[0161] Track the system effect in the long term, and work with business personnel to optimize algorithms and indicators to ensure continuous improvement. The implementation effect will be better if technical feasibility, business needs, cost control and other factors are comprehensively considered. It is also possible to start with a small-scale trial run and then gradually expand the scale.

[0162] Through the above-mentioned docking process with the business system, a closed-loop automatic advertising parameter adjustment system is built and continuously optimized to achieve the aforementioned technical effects.

[0163] The present invention also discloses a computer-readable storage medium, which stores program instructions executable by a processor to implement the following Figure 1 The steps of an embodiment of a method for automatically adjusting advertisement delivery parameters based on advertisement effects are shown.

[0164] The present invention also discloses a computer device, which comprises a memory, a processor and program instructions stored in the memory and executable by the processor, wherein the processor executes the program instructions to implement the following Figure 1 The steps of an embodiment of a method for automatically adjusting advertisement delivery parameters based on advertisement effects are shown.

[0165] Although the above methods are illustrated and described as a series of actions for simplicity of explanation, it should be understood and appreciated that these methods are not limited by the order of the actions, because according to one or more embodiments, some actions may occur in a different order and / or concurrently with other actions from those illustrated and described herein or not illustrated and described herein but understandable to those skilled in the art.

[0166] Those skilled in the art will further appreciate that the various illustrative logic blocks, modules, circuits, and algorithm steps described in conjunction with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or a combination of the two. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps are generally described above in terms of their functionality. Whether such functionality is implemented as hardware or software depends on the specific application and the design constraints imposed on the overall system. The technician may implement the described functionality in different ways for each specific application, but such implementation decisions should not be interpreted as resulting in a departure from the scope of the present invention.

[0167] The various illustrative logic blocks, modules, and circuits described in conjunction with the embodiments disclosed herein may be implemented or performed with a general purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general purpose processor may be a microprocessor, but in the alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. The processor may also be implemented as a combination of computing devices, such as a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in cooperation with a DSP core, or any other such configuration.

[0168] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. The software module may reside in a RAM memory, a flash memory, a ROM memory, an EPROM memory, an EEPROM memory, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium is coupled to a processor so that the processor can read and write information from / to the storage medium. In an alternative, a storage medium may be integrated into a processor. The processor and the storage medium may reside in an ASIC. The ASIC may reside in a user terminal. In an alternative, the processor and the storage medium may reside in a user terminal as discrete components.

[0169] In one or more exemplary embodiments, the functions described may be implemented in hardware, software, firmware, or any combination thereof. If implemented as a computer program product in software, each function may be stored on or transmitted by a computer-readable medium as one or more instructions or codes. Computer-readable media include both computer storage media and communication media, including any medium that facilitates the transfer of a computer program from one place to another. Storage media may be any available medium that can be accessed by a computer. As an example and not limitation, such a computer-readable medium may include RAM, ROM, EEPROM, CD-ROM or other optical disk storage, disk storage or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of an instruction or data structure and can be accessed by a computer. Any connection is also properly referred to as a computer-readable medium. For example, if the software is transmitted from a website, a server, or other remote source using a coaxial cable, a fiber optic cable, a twisted pair, a digital subscriber line (DSL), or wireless technologies such as infrared, radio, and microwaves, the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwaves are included in the definition of the medium. Disk and disc as used herein include compact disc (CD), laser disc, optical disc, digital versatile disc (DVD), floppy disk and Blu-ray disc, wherein disk often reproduces data magnetically, while disc reproduces data optically with lasers. Combinations of the above should also be included within the scope of computer-readable media.

[0170] The previous description of the disclosure is provided to enable any person skilled in the art to make or use the disclosure. Various modifications to the disclosure will be apparent to those skilled in the art, and the general principles defined herein may be applied to other variations without departing from the spirit or scope of the disclosure. Thus, the disclosure is not intended to be limited to the examples and designs described herein, but should be granted the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for automatically adjusting advertisement delivery parameters based on advertisement effect, characterized in that: Methods include: Step 1: Data preprocessing, including data cleaning, data conversion, and constructing feature vectors for advertising effectiveness indicators and subsequent training models; Step 2: training a decision model based on a deep learning network, wherein the input of the decision model is a feature vector of different parameter combinations, and the output of the decision model is a new advertising parameter combination and a predicted value of an advertising effect indicator; Step 3: Use the optimization strategy to obtain the predicted values ​​of the advertising effect indicators corresponding to each parameter combination in the optimization strategy based on the judgment model, and select the parameter combination corresponding to the highest value; Step 4: Use the parameter combination corresponding to the highest value in the advertising effect prediction value to actually place the advertisement, and collect the actual value of the advertising effect index of the placed advertisement; Step 5: Input the actual value of the advertising effect index obtained in step 4 into the judgment model, retrain the judgment model in an iterative manner, and save the best parameter combination of the advertising effect index after iterative training.

2. The method for automatically adjusting advertisement delivery parameters based on advertisement effect according to claim 1, characterized in that: Data cleaning includes removing duplicate records and outliers, and processing missing values; data conversion includes converting categorical variables into numerical features and standardizing them so that the feature values ​​are on the same scale; advertising effectiveness indicators include click-through rate, conversion rate, site rate, visit rate, or return on investment; the construction of feature vectors includes first constructing a single feature and then constructing interaction terms between single features; feature vectors include: advertising copy, advertising platform, advertising material, delivery time, target population, advertising expenditure, display volume, click volume, visit volume, conversion volume and time series data.

3. The method for automatically adjusting advertisement delivery parameters based on advertisement effect according to claim 1, characterized in that: In step 2, the deep learning network includes a long short-term memory network, a convolutional neural network, or a multi-layer perceptron.

4. The method for automatically adjusting advertisement delivery parameters based on advertisement effect according to claim 1, characterized in that: In step 3, the optimization strategy includes a random strategy or a genetic strategy.

5. A system for automatically adjusting advertisement delivery parameters based on advertisement effects, characterized in that: The system includes: Data preprocessing module, used for data preprocessing, including data cleaning, data conversion, and constructing feature vectors for advertising effect indicators and subsequent training models; A model training module is used to train a decision model based on a deep learning network, wherein the input of the decision model is a feature vector of different parameter combinations, and the output of the decision model is a new advertising parameter combination and a predicted value of an advertising effect indicator; The optimization module uses the optimization strategy to obtain the predicted values ​​of the advertising effect indicators corresponding to each parameter combination in the optimization strategy based on the judgment model, and selects the parameter combination corresponding to the highest value; The actual delivery module uses the parameter combination corresponding to the highest value in the advertising effect prediction value to actually deliver the advertisement and collect the actual value of the advertising effect index of the delivered advertisement; The iterative training module inputs the actual values ​​of the advertising effect indicators obtained by the actual delivery module into the judgment model, retrains the judgment model in an iterative manner, and saves the best parameter combination of the advertising effect indicators after iterative training.

6. The system for automatically adjusting advertisement delivery parameters based on advertisement effect according to claim 5, characterized in that: In the data preprocessing module, data cleaning includes removing duplicate records and outliers, and processing missing values; data conversion includes converting categorical variables into numerical features and standardizing them so that the feature values ​​are on the same scale; advertising effectiveness indicators include click-through rate, conversion rate, site rate, visit rate, or return on investment; the construction of feature vectors includes first constructing a single feature and then constructing interaction terms between single features; feature vectors include: advertising copy, advertising platform, advertising material, delivery time, target population, advertising expenditure, display volume, click volume, visit volume, conversion volume and time series data.

7. The system for automatically adjusting advertisement delivery parameters based on advertisement effect according to claim 5, characterized in that: In the model training module, the deep learning network includes long short-term memory network, convolutional neural network, or multi-layer perceptron.

8. The system for automatically adjusting advertisement delivery parameters based on advertisement effect according to claim 5, characterized in that: In the optimization module, the optimization strategy includes a random strategy or a genetic strategy.

9. A computer-readable storage medium storing program instructions executable by a processor to implement the steps of the method for automatically adjusting advertisement delivery parameters based on advertisement effects as described in any one of claims 1 to 4.

10. A computer device comprising a memory, a processor and program instructions stored in the memory and executable by the processor, wherein the processor executes the program instructions to implement the steps of the method for automatically adjusting advertisement delivery parameters based on advertisement effects as described in any one of claims 1 to 4.