Research on App Advertising Click Rate Evaluation Based on

Machine Learning Hybrid Model

Huaicang Li, Guochang Ma and Guodong Ma

Kunming University of Science and Technology, Yunnan Medical Health College, YunNan, 650000, China

Keywords: Click-Through Rate, Machine Learning, Hybrid Models, In App Advertising, Advertising Click-Through

Rate, Forecast.

Abstract: The role of click-through rate in App advertising is very important, but there are problems such as inaccurate

targeting of advertising content and low click-through rate. Traditional data analysis cannot solve the problem

of insufficient click-through rate and viewership in App ads, and the evaluation is unreasonable. Therefore,

this paper proposes a hybrid model of machine learning for advertising click-through rate prediction analysis.

Firstly, the social learning theory is used to evaluate the advertising content, and the indicators are divided

according to the advertising rating requirements to reduce the advertising rating Disturbing factors in . Then,

social learning theory evaluates the prediction and evaluation of the click-through rate of App advertising,

forms an evaluation scheme for the click-through rate of App advertising, and comprehensively analyzes the

click-through rate results. MATLAB simulation shows that under certain evaluation criteria, the machine

learning hybrid model is better than traditional data analysis in predicting the click-through rate and

viewership of App ads.

1 INTRODUCTION

Consumer satisfaction rate is one of the important

contents of App advertising click-through rate, which

is of great significance for advertising content

optimization (Li, and Wu, et al. 2023). However, in

the process of predicting views, the app advertising

click-through rate measurement scheme has the

problem of poor accuracy (Luo and Zhou, et al.

2023), which reduces the desire of app users to click

on ads. Some scholars believe that applying the

hybrid model of machine learning to the analysis of

insufficient click-through rate of advertisements can

effectively analyze the CTR evaluation scheme of

App advertising and provide corresponding support

for predicting page views (Onie and Berlinquette, et

al. 2023). On this basis, this paper proposes a hybrid

machine learning model to optimize the App

advertising click-through rate evaluation scheme

(Sahllal, and Souidi, 2023), and verifies the

effectiveness of the model.

Predicting the click-through rate of APP ads has

become a key task for advertisers and advertising

platforms (Sun, and Li, et al. 2023). To improve

prediction accuracy, researchers have come up with a

number of machine learning algorithms. This paper

will focus on the prediction of machine learning

hybrid algorithm on the click-through rate of APP ads

and the role of APP CTR (Tan and Bandyopadhyay, ,

et al. 2023).

1.1 App Ad Click-Through Rate

Prediction

1.1.1 Machine Learning Algorithms

Machine learning algorithms are the core technology

for ad click-through rate prediction. Among the many

machine learning algorithms, commonly used include

decision trees, naïve Bayes, support vector machines,

random forests, neural networks, etc (Wang and

Jiang, et al. 2023). These algorithms have their own

advantages and applicable scenarios, and enterprises

can choose the appropriate algorithm according to

their actual situation (Wang, and Yin, et al. 2023).

1.1.2 Data Preprocessing

Machine learning algorithms require large amounts of

data to support. In terms of data preprocessing, data

needs to be collected, cleaned, integrated and

transformed (Yan, and Li, et al. 2023). Data

Li, H., Ma, G. and Ma, G.

Research on App Advertising Click Rate Evaluation Based on Machine Learning Hybrid Model.

DOI: 10.5220/0013545100004664

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 3rd International Conference on Futuristic Technology (INCOFT 2025) - Volume 1, pages 435-440

ISBN: 978-989-758-763-4

435

collection refers to obtaining data from different

channels; data cleaning is to denoise, deduplicate and

correct data; data collation refers to the integration of

data in a specific format; and data transformation is

to convert data into a format acceptable to machine

learning algorithms (Yu and Ponomarenko, et al.

2023).

1.1.3 Feature Selection

Feature selection refers to the selection of data

features to extract the most meaningful features for

advertising click-through rate prediction. In terms of

feature selection, you need to select those

characteristics related to the click-through rate of

your ad and remove the invalid features. Commonly

used feature selection methods include information

gain, chi-square test, mutual information, etc (Yuan,

and Xu, et al. 2023).

1.1.4 Model Training

Model training refers to the input of preprocessed

data into a machine learning algorithm for training to

generate a predictive model in order to make

predictions on new data. The key to model training is

to select the appropriate algorithm and parameters,

and adjust the model to improve the prediction

accuracy (Zhang, and Han, et al. 2023).

1.2 APP Click-Through Rate Effect

1.2.1 Optimize Your Ad Serving Strategy

APP CTR can help ad delivery platforms optimize

advertising strategies. According to the level of the

APP click-through rate, you can decide whether to

stop the delivery of certain ads, adjust the advertising

period, adjust the placement of advertisements, etc.,

to improve the exposure and click-through rate of

advertisements.

1.2.2 Evaluate Ad Performance

App CTR can evaluate the effectiveness of ads.

Through the statistics and analysis of advertising

display and click data, you can evaluate the

effectiveness and attractiveness of ads, and then

adjust the display and content of advertisements to

improve the click-through rate and conversion rate of

ads.

1.2.3 Applied to Recommender Systems

APP CTR can also be applied to recommender

systems. Recommendation system refers to

recommending products or services that meet the

needs of users based on their behavior and interests.

APP click-through rate can be used as an important

indicator of the recommendation system, accurately

predict the user's interest and demand for a certain

product or service, and improve the accuracy and user

experience of the recommendation system.

1.2.4 Optimize the User Experience

The app click-through rate can reflect the user

experience to a certain extent. If the click-through rate

of your ad is too low, it means that the content of the

ad does not match the user's interests, and the content

and display method need to be optimized to improve

the user experience. Through the analysis and

feedback of the APP click-through rate, the user

experience can be optimized and user satisfaction and

loyalty can be improved.

1.3 Learn Hybrid Algorithms

Machine learning hybrid algorithms refer to the

combination of multiple different machine learning

algorithms and the combined use of their advantages

to improve prediction accuracy. Commonly used

machine learning hybrid algorithms include Bagging,

Boosting, Stacking, etc.

1.3.1 Bagging

Bagging, short for Bootstrap aggregating, is an

ensemble learning method based on autonomous

resampling. Bagging samples the data with put back,

generates multiple datasets, then trains the same

model for each dataset, and finally averages or votes

on the results of multiple models to generate the final

prediction results. The advantage of Bagging is that it

can reduce the variance of the model and improve the

prediction accuracy.

1.3.2 Boosting

Boosting is a method of gradually improving the

accuracy of a model. Boosting samples the data back,

generates multiple datasets, then trains each dataset

with different models, and finally weights the results

of multiple models to generate the final prediction

results. The advantage of Boosting is that it can

reduce the bias of the model and improve the

prediction accuracy.

INCOFT 2025 - International Conference on Futuristic Technology

436

1.3.3 Stacking

Stacking is a machine learning approach based on

model integration. Stacking samples the data with

placement, generates multiple datasets, and then

inputs each dataset into a number of different models

for training, and finally inputs the results of multiple

models into another model for training to generate the

final prediction results. The advantage of stacking is

that it can combine the advantages of different models

to improve prediction accuracy.

APP advertising click-through rate prediction is

an important task for advertising platforms, and

machine learning hybrid algorithms can improve the

accuracy of advertising click-through rate prediction.

APP CTR can help advertising platforms optimize

advertising strategies, evaluate advertising effects,

apply to recommendation systems, and optimize user

experience. Companies should choose the right

machine learning algorithm and hybrid algorithm

according to their actual needs to improve the

accuracy and precision of ad click-through rate

prediction.

2 RELATED CONCEPTS

2.1 Mathematical Description of a

Machine Learning Hybrid Model

The machine learning hybrid model uses big data

theory to optimize the app advertising click-through

rate evaluation scheme, and finds the unqualified

values in the app advertisement according to the

indicators in the predicted page views, and integrates

the app advertising click-through rate evaluation

scheme to finally judge the feasibility of insufficient

advertising click-through rate. The hybrid machine

learning model combines the advantages of big data

theory to quantify the lack of click-through rate of

advertisements, which can improve advertising

viewership and consumer satisfaction.

Hypothesis 1: The advertising rating requirement

, that the App advertising click-through rate

evaluation scheme is

set , the satisfaction of the App

advertising click-through rate measurement scheme

, and the judgment function of the App

advertising click-through rate evaluation scheme is

(0)

Lx≈ ,as shown in Equation (1).

(r ) (r ) 4

ii i

Lset X

=→ −⋅

 

)

2.2 Choice of Consumer Satisfaction

Rate Program

Hypothesis 2: The ad click-through rate function is

and the weight factor is

K( )

, then the ad ratings

requirement is

, the click-through rate of the

unqualified ad as shown in Equation (2).

K( )= lim

ii i

xd dx

→

Ω−

⋅⋅⋅→⋅R

)

2.3 Analysis of App Advertising Click-

Through Rate Measurement

Scheme

Before the machine learning hybrid model, multi-

dimensional analysis should be carried out on the App

advertising click-through rate evaluation scheme, and

the advertising rating requirements should be mapped

to the advertising click-through rate insufficient

library, and the unqualified App advertising click-

through rate evaluation scheme should be eliminated.

First, a comprehensive analysis of the insufficient

click-through rate of advertisements is carried out,

and the threshold and index weight of the App

advertising click-through rate evaluation scheme are

set to ensure the accuracy of the hybrid model of

machine learning. The advertising click-through rate

is not enough to test the app advertising click-through

rate measurement scheme for the system, and

standardized analysis is required. If the

advertisement's CTR is insufficient in a non-normal

distribution, its App ADR measurement scheme will

be affected, reducing the accuracy of the overall ad

viewership. In order to improve the accuracy of the

machine learning hybrid model and improve the

optimization of advertising ratings, it is necessary to

select the App advertising click-through rate

measurement scheme, and the specific scheme

selection is shown in Figure 1.

The survey App advertising click-through rate

evaluation scheme shows that the consumer

satisfaction rate scheme presents a multi-dimensional

distribution, which is in line with objective facts. The

insufficient click-through rate of advertising is not

Research on App Advertising Click Rate Evaluation Based on Machine Learning Hybrid Model

437

directional, indicating that the consumer satisfaction

rate scheme has strong randomness, so it is regarded

as a high analytical study. The insufficient advertising

click-through rate meets the normal requirements,

mainly because the big data theory adjusts the

insufficient advertising click-through rate, removes

the duplicate and irrelevant schemes, and

supplements the default scheme, so that the dynamic

correlation of the entire App advertising click-

through rate measurement scheme is strong.

Click-through rate

Number of

advertisements

Advertising time

content

Consumer

Figure 1: Consumer Satisfaction Program Selection Results

3 OPTIMIZATION STRATEGIES

FOR INSUFFICIENT

ADVERTISING CLICK-

THROUGH RATE

The machine learning hybrid model adopts a random

optimization strategy for insufficient click-through

rate of ads, and adjusts the parameters of advertising

content to optimize the scheme of insufficient click-

through rate of ads. The machine learning hybrid

model divides the insufficient click-through rate of

advertisements into different ad rating levels, and

randomly selects different schemes. In the iterative

process, the CTR evaluation scheme of App ads with

different ad rating levels is optimized and analyzed.

After the optimization analysis is completed, compare

the ad viewership optimization of different scenarios

and record that the best ad click-through rate is

insufficient.

4 ACTUAL CASES OF

INSUFFICIENT CLICK-

THROUGH RATE FOR

ADVERTISING

4.1 Advertising Ratings Profile

In order to facilitate the analysis of advertising

ratings, this paper takes the insufficient advertising

click-through rate in complex situations as the

research object, with 12 paths and a test time of 12h

shown in Table 1.

Table 1: App ad viewership requirements

Scope of

application

Frequency Standard

effect

Consumer

satisfaction

rate

Juvenile 1~30,000 74.95 75.85

4~80,000 82.92 80.78

Adult 1~30,000 76.39 81.77

4~80,000 81.88 79.27

Midlife 1~30,000 79.06 79.04

4~80,000 78.19 80.05

The advertising viewership process in Table 1. is

shown in Figure 2.

Consumer

satisfaction

Inaccurate

orientation

Advertising

ratings

click rate

Traditio nal

data analysis

Figure 2: Analysis process of insufficient click-through rate

of advertisements

INCOFT 2025 - International Conference on Futuristic Technology

438

Compared with traditional data analysis, the app

ad click-through rate measurement scheme of

machine learning hybrid model is closer to the actual

prediction page view requirements. In terms of the

rationality and fluctuation range of insufficient

advertising click-through rate, machine learning

hybrid model traditional data analysis. It can be seen

from the change of the app advertising click-through

rate measurement scheme in Figure II that the

machine learning hybrid model has better stability

and faster judgment speed. Therefore, the App ad

click-through rate measurement scheme of machine

learning hybrid model has better speed, consumer

satisfaction rate scheme, App advertising click-

through rate measurement scheme, and summation

stability.

4.2 Insufficient Click-Through Rate of

Advertising

The CTR measurement scheme for App ads with

insufficient click-through rate includes unstructured

information, semi-structured information, and

structural information. After the pre-selection of

machine learning hybrid model, a preliminary click-

through rate evaluation scheme for App ads with

insufficient click-through rate was obtained, and the

feasibility of the click-through rate evaluation scheme

for App ads with insufficient click-through rate was

analyzed. In order to more accurately verify the

standard effect of insufficient click-through rate,

select the advertising click-through rate of different

advertising rating levels, and the App advertising

click-through rate evaluation scheme is shown in

Table 2.

Table 2: The overall picture of the consumer satisfaction

rate program

Cate

Satisfaction Anal

sis rate

Juvenile 90.55 90.81

Adult 90.41 85.63

Midlife 87.96 85.55

Mean 84.81 88.95

88.10 87.05

P=1.728

4.3 Consumer Satisfaction Rate and

Stability of Advertising Ratings

In order to verify the accuracy of the hybrid machine

learning model, compared with the traditional data

analysis of App advertising click-through rate

measurement scheme, the App advertising click-

through rate measurement scheme is shown in Figure

It can be seen from Figure 3 that the consumer

satisfaction rate of the machine learning hybrid model

is higher than that of traditional data analysis, but the

Figure 3: Consumer satisfaction rates for different

algorithms

error rate is lower, indicating that the advertising

rating of the machine learning hybrid model is

relatively stable, while the advertising rating of

traditional data analysis is uneven. The average app

ad click-through rate measurement scheme of the

above three algorithms is shown in Table 3.

Table 3: Comparison of ad viewership accuracy by different

methods

Algorithm Consumer

satisfaction

rate

Magnitude

of change

Error

Machine

learning

hybrid

models

91.54 89.93 89.70

Traditional

data analysis

89.12 90.01 87.47

P 89.60 88.69 89.89

It can be seen from Table 3 that traditional data

analysis has deficiencies in consumer satisfaction rate

and stability in terms of insufficient advertising click-

through rate, and the insufficient advertising click-

through rate has changed significantly, and the error

rate is high. The general results of machine learning

hybrid models have higher consumer satisfaction

rates and are better than traditional data analysis. At

the same time, the consumer satisfaction rate of the

machine learning hybrid model is greater than 88%,

and the optimization has not changed significantly. In

order to further verify the superiority of machine

learning hybrid models. In order to further verify the

effectiveness of the proposed method, the machine

Research on App Advertising Click Rate Evaluation Based on Machine Learning Hybrid Model

439

learning hybrid model is analyzed in general with

different methods, as shown in Figure 4.

Figure 4: Consumer satisfaction rate for machine learning

hybrid model ad ratings

It can be seen from Figure 4 that the consumer

satisfaction rate of the machine learning hybrid

model is significantly better than that of traditional

data analysis, and the reason is that the machine

learning hybrid model increases the adjustment factor

of insufficient advertising click-through rate, and sets

the threshold of advertising content to eliminate the

App advertising click-through rate evaluation scheme

that does not meet the requirements.

5 CONCLUSIONS

Aiming at the problem of insufficient click-through

rate of advertising and unsatisfactory consumer

satisfaction rate, this paper proposes a hybrid model

of machine learning, and combines big data theory to

optimize the insufficient click-through rate of

advertising. At the same time, in-depth analysis of

advertising rating standards and threshold standards

is carried out to construct advertising content

collection. The research shows that the machine

learning hybrid model can improve the optimization

and stability of insufficient advertising click-through

rate, and can conduct general viewership analysis on

insufficient advertising click-through rate. However,

in the process of machine learning hybrid model, too

much attention is paid to the analysis of advertising

ratings, resulting in irrationality in the selection of

advertising ratings indicators.

REFERENCES

Li, Y., Wu, R., Liu, J., & Wang, S.(2023) The Impact of

Targeted Online Advertising's Pushing Time on

Consumers' Browsing Intention: A Study Based on

Regret Theory. Journal of Global Information

Management, 31(1):13

Luo, H., Zhou, X., Ding, H., & Wang, L.(2023) Multi-Task

Deep Learning with Task Attention for Post-Click

Conversion Rate Prediction. Intelligent Automation

and Soft Computing, 36(3): 3583-3593.

Onie, S., Berlinquette, P., Holland, S., Livingstone, N.,

Finemore, C., Gale, N., Elder, E., Laggis, G.,

Heffernan, C., Armstrong, S. O., Theobald, A.,

Josifovski, N., Torok, M., Shand, F., & Larsen,

M.(2023) Suicide Prevention Using Google Ads:

Randomized Controlled Trial Measuring Engagement.

Jmir Mental Health, 10(3):3.

Sahllal, N., & Souidi, E. M.(2023) A Comparative Analysis

of Sampling Techniques for Click-Through Rate

Prediction in Native Advertising. Ieee Access,

11(3):24511-24526.

Sun, X., Li, Y., Guo, B., & Gao, L.(2023) Marketing

Automation: How to Effectively Lead the Advertising

Promotion for Social Reconstruction in Hotels.

Sustainability, 15(5):89.

Tan, S. Z., Bandyopadhyay, A., & Septianto, F.(2023)

Relationship (breakup) reminders drive online

advertising effectiveness. Psychology & Marketing,

40(6): 1152-1161.

Wang, P., Jiang, L., & Yang, J.(2023) The Early Impact of

GDPR Compliance on Display Advertising: The Case

of an Ad Publisher. Journal of Marketing

Research,1(3):102.

Wang, Y., Yin, H., Wu, L., Chen, T., & Liu, C.(2023)

Secure Your Ride: Real-Time Matching Success Rate

Prediction for Passenger-Driver Pairs. Ieee

Transactions on Knowledge and Data Engineering,

35(3): 3059-3071.

Yan, C., Li, X., Zhang, Y., Wang, Z., & Wan, Y.(2023)

MIN: multi-dimensional interest network for click-

through rate prediction. Knowledge and Information

Systems,7(3):89.

Yu, Z., Ponomarenko, V., & Liska, L. I.(2023) How to

Allocate White Space in Ad Design? The Impact of

Product Layouts on Perceived Entitativity and

Advertising Performance. Journal of

Advertising,6(3):18-22.

Yuan, Y., Xu, F., Cao, H., Zhang, G., Hui, P., Li, Y., & Jin,

D.(2023) Persuade to Click: Context-Aware Persuasion

Model for Online Textual Advertisement. Ieee

Transactions on Knowledge and Data Engineering,

35(2): 1938-1951.

Zhang, W., Han, Y., Yi, B., & Zhang, Z.(2023) Click-

through rate prediction model integrating user interest

and multi-head attention mechanism. Journal of Big

Data, 10(1):56.

INCOFT 2025 - International Conference on Futuristic Technology

440