KR-20260065732-A - SERVER, METHOD, AND COMPUTER PROGRAM FOR PREDICTING ADVERTISING PERFORMANCE BASED ON SEASONALITY DATA AND EVENT DATA

KR20260065732AKR 20260065732 AKR20260065732 AKR 20260065732AKR-20260065732-A

Abstract

A server providing a service for predicting advertising performance comprises: a receiving unit that receives advertising performance data, seasonal data, and event data from at least one advertising channel; a learning unit that preprocesses the advertising performance data, seasonal data, and event data, and trains a plurality of deep learning models based on the preprocessed advertising performance data, seasonal data, and event data; a prediction unit that predicts and provides a result value for an advertising performance indicator, including any one of Click-Through Rate (CTR), Conversion Rate (CVR), Return on Ad Spend (ROAS), and Cost Per Action (CPA), by performing mixed inference from the plurality of deep learning models based on seasonal factors and/or event factors; and a recommendation unit that post-processes the predicted result value for the advertising performance indicator and generates and provides advertising strategy recommendation data based on the post-processed predicted result value. and includes an execution unit that performs advertising execution for at least one advertising channel based on the advertising strategy recommendation data, wherein the plurality of deep learning models includes any one of an ARIMA (Autoregressive Integrated Moving Average) model, a SARIMA (Seasonal ARIMA) model, an LSTM (Long Short-Term Memory) model, and a Transformer model.

Inventors

이병갑
모선영
김도수

Assignees

애디플 주식회사

Dates

Publication Date: 20260511
Application Date: 20251031
Priority Date: 20241101

Claims (19)

In a server that provides a service for predicting advertising performance, A receiver that receives advertising performance data, seasonal data, and event data from at least one advertising channel; A learning unit that preprocesses the above advertising performance data, the above seasonal data, and the above event data, and trains a plurality of deep learning models based on the preprocessed advertising performance data, seasonal data, and event data; A prediction unit that predicts and provides a result value for an advertising performance indicator, including any one of Click-Through Rate (CTR), Conversion Rate (CVR), Return on Ad Spend (ROAS), and Cost Per Action (CPA), based on seasonal factors and/or event factors, by performing mixed inference from the above plurality of deep learning models; A recommendation unit that performs post-processing on the predicted result value for the above-mentioned advertising performance indicator and generates and provides advertising strategy recommendation data based on the post-processed predicted result value; and An execution unit that performs advertising execution for at least one of the above advertising channels based on the above advertising strategy recommendation data Includes, An advertising performance prediction server in which the plurality of deep learning models includes any one of an ARIMA (Autoregressive Integrated Moving Average) model, a SARIMA (Seasonal ARIMA) model, an LSTM (Long Short-Term Memory) model, and a Transformer model.
In Article 1, The above learning unit is, Generate each prediction value for the above ARIMA model, the above SARIMA model, the above LSTM model, and the above Transformer model, and metamodel the prediction values, An advertising performance prediction server based on the MultiLayer Perceptron (MLP) and XGBoost models as metamodels.
In Article 2, The above learning unit is, Daily advertising performance data is input into the above ARIMA model to output predicted short-term advertising performance values, and By inputting daily advertising performance data and a preset seasonal cycle into the above SARIMA model, an advertising performance value reflecting predicted seasonal factors is output, and The above LSTM model is input with advertising performance indicators, event flags, seasonal data, and event data for a preset period to output predicted advertising performance values for a future preset period, and An advertising performance prediction server that inputs event sequences, time encodings, and external variable data into the above-mentioned Transformer model to output predicted event impact.
In Paragraph 3, The above learning unit is, An advertising performance prediction server that inputs the short-term advertising performance value, the advertising performance value reflecting seasonal factors, the advertising performance value for a future preset period, and the event impact degree into the metamodel to provide a predicted result value for the final advertising performance indicator.
In Article 1, The above prediction unit is, An advertising performance prediction server that assigns weights to the plurality of deep learning models using a Mixture-of-Experts (MoE) model.
In Article 5, The above prediction unit is, An advertising performance prediction server in which a gating network for the above MoE model assigns the above weights based on the context included in each input data.
In Article 1, The above recommendation section is, An advertising performance prediction server that generates and provides the above-mentioned advertising strategy recommendation data as visualized data.
In Article 7, An advertising performance prediction server in which the above-mentioned visualized data includes any one of the following: a predicted result value for the above-mentioned advertising performance indicator, information suggesting advertising channels and creatives, and information suggesting the optimal timing for advertising and an advertising budget plan.
In Article 1, The aforementioned executive body, An advertising performance prediction server that redistributes the advertising budget or optimizes targeting and creatives based on the above advertising strategy recommendation data.
In a method for providing a service that predicts advertising performance, A step of receiving advertising performance data, seasonal data, and event data from at least one advertising channel; A step of preprocessing the above advertising performance data, the above seasonal data, and the above event data, and training a plurality of deep learning models based on the preprocessed advertising performance data, seasonal data, and event data; A step of predicting and providing a result value for an advertising performance indicator, including any one of Click-Through Rate (CTR), Conversion Rate (CVR), Return on Ad Spend (ROAS), and Cost Per Action (CPA), based on seasonal factors and/or event factors, by performing mixed inference from the above plurality of deep learning models; A step of post-processing the predicted result value for the above-mentioned advertising performance indicator, and generating and providing advertising strategy recommendation data based on the post-processed predicted result value; and Step of executing an advertisement for at least one of the above advertising channels based on the above advertising strategy recommendation data Includes, An advertising performance prediction method in which the above plurality of deep learning models includes any one of an ARIMA (Autoregressive Integrated Moving Average) model, a SARIMA (Seasonal ARIMA) model, an LSTM (Long Short-Term Memory) model, and a Transformer model.
In Article 10, The step of training the above plurality of deep learning models is, A step of generating each prediction value for the above ARIMA model, the above SARIMA model, the above LSTM model and the above Transformer model; and Step of metamodeling the above predicted values Includes, An advertising performance prediction method based on the MultiLayer Perceptron (MLP) and XGBoost models as metamodels.
In Article 11, The step of training the above plurality of deep learning models is, A step of inputting daily advertising performance data into the above ARIMA model to output predicted short-term advertising performance values; A step of inputting daily advertising performance data and a preset seasonal cycle into the above SARIMA model to output advertising performance values reflecting predicted seasonal factors; A step of inputting advertising performance indicators, event flags, seasonal data, and event data for a preset period into the above LSTM model to output predicted advertising performance values for a future preset period; and A step of inputting event sequence, time encoding, and external variable data into the above Transformer model to output the predicted event impact. An advertising performance prediction method that further includes
In Article 12, The step of training the above plurality of deep learning models is, A step of providing a predicted result value for the final advertising performance indicator by inputting the short-term advertising performance value, the advertising performance value reflecting seasonal factors, the advertising performance value for a future preset period, and the event impact into the metamodel. An advertising performance prediction method that further includes
In Article 10, The step of predicting and providing result values for the above-mentioned advertising performance indicators is: A step of assigning weights to the plurality of deep learning models using a Mixture-of-Experts (MoE) model. An advertising performance prediction method that further includes
In Article 14, The step of predicting and providing result values for the above-mentioned advertising performance indicators is: A step in which the gating network for the above MoE model assigns the weights based on the context contained in each input data. An advertising performance prediction method that further includes
In Article 10, The step of generating and providing the above-mentioned advertising strategy recommendation data is, Step of generating and providing the above advertising strategy recommendation data as visualized data An advertising performance prediction method that further includes
In Article 16, An advertising performance prediction method in which the above-mentioned visualized data includes any one of the following: a predicted result value for the above-mentioned advertising performance indicator, information suggesting advertising channels and creatives, and information suggesting the optimal timing for advertising and an advertising budget plan.
In Article 10, The step of executing the above advertisement is, Steps to redistribute the advertising budget or optimize targeting and creatives based on the above advertising strategy recommendation data. An advertising performance prediction method that further includes
In a computer program stored on a computer-readable recording medium comprising a sequence of instructions that provide a service for predicting advertising performance, When the above computer program is executed by a computing device, Receive advertising performance data, seasonal data, and event data from at least one advertising channel, and The above advertising performance data, the above seasonal data, and the above event data are preprocessed, and a plurality of deep learning models are trained based on the preprocessed advertising performance data, seasonal data, and event data, and By performing blended inference from the aforementioned multiple deep learning models, a result value for an advertising performance indicator including any one of Click-Through Rate (CTR), Conversion Rate (CVR), Return on Ad Spend (ROAS), and Cost Per Action (CPA) based on seasonal and/or event factors is predicted and provided. Post-processing is performed on the predicted results for the above advertising performance indicators, and advertising strategy recommendation data is generated and provided based on the post-processed predicted results. Based on the above advertising strategy recommendation data, advertising execution for at least one of the above advertising channels is performed, A computer program stored on a computer-readable recording medium, wherein the plurality of deep learning models comprises a sequence of instructions including any one of an ARIMA (Autoregressive Integrated Moving Average) model, a SARIMA (Seasonal ARIMA) model, an LSTM (Long Short-Term Memory) model, and a Transformer model.

Description

Server, method, and computer program for predicting advertising performance based on seasonality data and event data The present invention relates to a server, a method, and a computer program for predicting the performance of advertisements based on seasonal data and event data. Advertising performance prediction technology is the process of forecasting future advertising results through data analysis. To this end, various Key Performance Indicators (KPIs) are analyzed, and if performance is poor, issues such as product detail pages, keywords, and bidding strategies are diagnosed. Recently, solutions utilizing AI technology to assist in optimal budget allocation and campaign strategy formulation have also emerged. However, conventional advertising performance prediction technologies tend to rely on static models (Static Training/Static Computational Graph), which has the disadvantage of failing to reflect variability caused by seasonal or event-based factors. Furthermore, conventional advertising performance prediction technologies have limitations in processing statistically non-linear patterns and large-scale data, and automation is restricted to a single advertising channel, resulting in significant operational complexity and resource waste when managing multiple channels. Figure 1 is a configuration diagram of an advertising performance prediction server. Figure 2 is an exemplary diagram of multiple deep learning models. Figure 3a is an exemplary diagram of visualized data. Figure 3b is an exemplary diagram of visualized data. Figure 4 is a flowchart of the advertising performance prediction method. Embodiments of the present invention are described below with reference to the attached drawings so that those skilled in the art can easily implement the invention. However, the present invention may be embodied in various different forms and is not limited to the embodiments described herein. Furthermore, in order to clearly explain the present invention in the drawings, parts unrelated to the explanation have been omitted, and similar parts throughout the specification are denoted by similar reference numerals. Throughout the specification, when a part is described as being "connected" to another part, this includes not only cases where they are "directly connected" but also cases where they are "electrically connected" with other elements interposed between them. Furthermore, when a part is described as "including" a component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components, and it should be understood that this does not preclude the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof. In this specification, the term "part" includes a unit realized by hardware, a unit realized by software, and a unit realized using both. Additionally, one unit may be realized using two or more hardware, and two or more units may be realized by one hardware. Some of the operations or functions described in this specification as being performed by a terminal or device may instead be performed by a server connected to said terminal or device. Likewise, some of the operations or functions described as being performed by a server may also be performed by a terminal or device connected to said server. An embodiment of the present invention will be described in detail below with reference to the attached drawings. FIG. 1 is a configuration diagram of an advertising performance prediction server according to an embodiment of the present invention. Referring to FIG. 1, the advertising performance prediction server (100) may include a receiving unit (110), a learning unit (120), a prediction unit (130), a recommendation unit (140), and an execution unit (150). However, the above components (110 to 150) are merely illustrative examples of components that can be controlled by the advertising performance prediction server (100). Each component of the advertising performance prediction server (100) of FIG. 1 is generally connected via a network. For example, as shown in FIG. 1, the receiving unit (110), the learning unit (120), the prediction unit (130), the recommendation unit (140), and the execution unit (150) may be connected simultaneously or at intervals. A network refers to a connection structure capable of exchanging information among respective nodes, such as terminals and servers, and includes local area networks (LAN), wide area networks (WAN), the internet (WWW: World Wide Web), wired and wireless data communication networks, telephone networks, wired and wireless television communication networks, etc. Examples of wireless data communication networks include, but are not limited to, 3G, 4G, 5G, 3GPP (3rd Generation Partnership Project), LTE (Long Term Evolution), WIMAX (World Interoperability for Microwave Access), Wi-Fi, Bluetooth communication, infrared communication, u