Application and Optimization of Cloud Computing in Big Data

Processing

Zhong Chen

, Guoyan Yang and Feijiang Huang

School of Information and Communication Engineering,

Guangzhou Maritime University, Guangzhou, Guangdong, 510725, China

Keywords: Cloud Computing, Big Data Processing, E-Commerce Platforms, Application & Optimization.

Abstract: This paper studies the application of cloud computing in big data processing, taking the user behavior analysis

of e-commerce platform as an example, aiming at the recommendation accuracy and user conversion rate of

the platform. Based on the collection of user operation, product sales and regional data, the relationship

between user behavior and conversion rate is analyzed. In this paper, it is necessary to build the platform

architecture and framework, optimize the framework, and apply it, and finally conclude that the dwell time of

purchase behavior is highly correlated with the conversion rate, and the high unit price products contribute

significantly to sales. After analyzing the experimental data, it was concluded that the conversion rate was the

highest in the northern region, while the conversion rate was lower in the southern region despite the high

number of visits. Based on the comprehensive analysis, it was concluded that improving user dwell time and

optimizing regional strategies could help improve its overall sales. This also shows that the application of

cloud computing in big data processing is very effective, and effective improvements can be achieved through

intelligent algorithm optimization.

1 INTRODUCTION

Because of the rapid development of e-commerce

platforms, the analysis of user behavior data has

become a key factor in improving their conversion

rate (Ahmad, and Khan, et al. 2023). Many

researchers have proposed that user conversion can be

optimized based on recommendation algorithms, but

the above methods do not perform well in processing

big data, and cannot effectively deal with high-

dimensional features and complex user behaviors.

Some researchers have proposed a method based on

improved recommendation algorithms, but the

computing resources of the above methods are very

efficient and cannot be applied on a large scale

(Alshareef, 2023). This paper uses the combination of

cloud computing technology and intelligent

algorithms to analyze user behavior by using the

powerful computing power of cloud computing

(Arulmozhiselvan, and Uma, 2024). Through

distributed processing, this paper hopes to improve

the accuracy of platform recommendations. Another

reason for choosing this method is that it can really

process large amounts of data and is extremely good

at predicting and optimizing user conversions in real

time (He, Y. J and W. H. Ouyang, et al. 2023).

2 RELATED WORKS

2.1 Cloud Computing Technology

Cloud computing technology is a technical system

that provides computing resources based on the

network, which is characterized by on-demand

allocation, scalability, and elastic computing. In a

cloud computing environment, users do not need to

manage physical hardware (Kang, and Deng, 2023),

but can cope with computing needs in different

periods as long as computing resources are

dynamically adjusted. This type of on-demand

allocation model is particularly suitable for large-

scale data processing, such as real-time user data

analysis on e-commerce platforms (Kumar, and Saini,

2024). Cloud computing is based on virtualization

technology, which can process massive data requests

Chen, Z., Yang, G. and Huang, F.

Application and Optimization of Cloud Computing in Big Data Processing.

DOI: 10.5220/0013546200004664

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 485-491

ISBN: 978-989-758-763-4

485

at the same time, ensuring the efficient operation of

the platform and improving resource utilization

(Muhic, and Bengtsson, et al. 2023).

2.2 Application of Big Data Processing

The theory of big data processing mainly includes the

storage, management, and analysis of massive data.

The theory needs to be implemented through a

distributed computing framework (Nagahawatta, and

Warren, et al. 2024). For example, Spark can

efficiently store and process data from different

sources (Pericherla, 2023). Big data is characterized

by large data volume, complex structure, and strong

real-time performance, so it is necessary to improve

processing efficiency based on parallel row

computing and distributed storage. Big data

processing also includes data mining techniques, such

as extracting valuable information from user behavior

to help platforms make personalized

recommendations and market decisions (Pham, and

Huynh-The, et al. 2023).

2.3 Mechanism of User Behavior

Analysis

The theory of user behavior analysis is based on a

detailed analysis of the various actions of users on the

platform to understand the interests and preferences

of users. Based on data mining and machine learning

algorithms, the platform will be able to extract user

behavior patterns from historical data, such as

browsing and clicking, adding shopping carts, and

purchasing behaviors. For example, user dwell time,

product browsing order, click-through rate, bounce

rate, etc. are all important parameters in behavioral

analysis. Based on the analysis of the above

parameters, e-commerce platforms can optimize the

recommendation platform and improve the purchase

conversion rate of its users, providing strong support

for platform operations.

3 THE REALIZATION OF

E-COMMERCE PLATFORMS

COMBINING CLOUD

COMPUTING AND BIG DATA

PROCESSING

3.1 The Architecture of the

E-Commerce User Behavior

Analysis Platform

In the study, it is necessary to build a platform,

because this article takes the analysis of user behavior

of e-commerce platforms as an example, so it needs

to have these blocks. It includes user interface blocks,

request processing blocks, data storage blocks, data

analysis blocks, recommendation engine blocks, and

monitoring and feedback blocks. Among them, the

user interface block is mainly responsible for

interacting with the user. Based on the webpage,

mobile terminal and other interfaces, it collects user

input, such as search, click, purchase and other

behaviors, and transmits user requests to the backend

of the platform to ensure the friendliness and

efficiency of the user experience. The request

processing block is mainly responsible for receiving

requests from the user interface block and scheduling

appropriate services according to the type of request.

Blocks are responsible for parsing user needs, such as

product searches, recommendation requests, and

interacting with back-end data processing blocks. The

data storage block is responsible for storing all user

data of the system, such as user profiles, browsing

history, and purchase history. It is based on a

distributed storage system to ensure the durability and

fast reading of data

And it supports real-time query. The data analysis

block is mainly responsible for real-time analysis of

stored data and extracting useful user behavior

patterns. Based on the analysis of users' browsing and

purchase behavior, personalized recommendation

results are generated, and data support is provided for

subsequent user predictions. The recommendation

engine block is mainly responsible for generating

personalized product or content recommendations

based on the historical behavior and current needs of

its users. It leverages a well-established framework

combined with real-time data to generate dynamic

recommendations to improve user satisfaction. The

monitoring and feedback block is mainly responsible

for monitoring the operation status of the platform in

real time, such as the response speed of user

interaction and the accuracy of recommendation

INCOFT 2025 - International Conference on Futuristic Technology

486

results. Based on the analysis of user feedback data,

the block can dynamically adjust the platform to

ensure the stability and optimization of the platform.

3.2 Construction of User Behavior

Analysis Framework

Data sampling and distributed storage, in which based

on data partitioning, cloud computing can distribute

transaction data from users around the world to

individual nodes for parallel computing, while

reducing the pressure on a single node and improving

processing efficiency. See Eq. (1) for details.

{

}

iii

DX,y =

(1)

In the formula,

is data separation, which

represents the partition of massive data by the e-

commerce user behavior analysis platform, such as

the user transaction records of the e-commerce

website. Each partition is assigned to a different node

for parallel processing, such as data retrieved from

servers in different regions. The partitioned data is

stored in parallel on platforms such as Hadoop.

Represents the feature set of the sample, which refers

to the characteristics of user behavior on the e-

commerce platform, such as the user's browsing time

and the type of product clicked. The above

characteristics will be used to predict the user's

purchase behavior. For example,

indicates a

purchase and 0 indicates that a user has not

purchased. The above labels can guide the generation

of decision trees when the framework is strengthened.

Subsequently, the selection of random features and

the optimization of cloud computing nodes were

carried out. See (2).

sub k

Ff,f,,f

=…

(2)

In the formula,

is all available characteristics

are represented, such as the user's click-through rate,

product category, device type, etc. These

characteristics help the framework understand user

behavior and provide personalized recommendations.

sub

Represents a randomly selected set of features

for each decision tree, such as browse time,

commodity price, etc. Each tree is built using only a

subset of these features, which increases the diversity

of the framework. Based on the parallel processing of

different feature sets at multiple nodes, the

framework can strengthen multiple trees at the same

time and shorten the strengthening time. In the cloud

computing environment, random feature selection

can help e-commerce platforms quickly process user

behavior data, such as analyzing whether users will

buy goods for certain characteristics, and improving

the response speed and accuracy of their

recommendation platforms.

However, the construction of decision trees and

parallel processing are carried out. Specifically, in a

distributed environment, e-commerce platforms can

process multiple decision trees in parallel to capture

different purchase patterns, such as different

behaviors of users when browsing, adding to carts,

and checking out. Based on distributed parallel

computing, it can quickly process huge data sets and

shorten prediction time. See Eq. (3) for this.

iisub

T(X,F)Tree=

)

In the formula,

refers to the first decision tree,

and each node processes specific data partitions and

feature subsets in parallel to build a decision tree. In

practice, each tree can capture different user behavior

patterns, with some trees focusing on user browsing

time and others focusing on price sensitivity.

Refers to user behavior data, such as features

extracted from the order in which users visit the e-

commerce platform. These features are used to

enhance the decision tree and predict whether the user

will buy the product or not.

3.3 Framework Enhancement and

Optimization

In order to ensure the further optimization of the

framework, it needs to be continuously strengthened

until the iteration is completed and its performance in

all aspects meets the requirements. The targeted

reinforcement of the framework can improve the

adaptability of the framework to different

environments as much as possible, so as to better

improve the practical application performance of the

framework. In the process of reinforcement, it is

necessary to optimize the number of trees and

resource scheduling, and based on this, better

reinforcement is carried out to meet the final

requirements. See Eq. (4) for details.

NargminL(y,F(X;N))=

)

Application and Optimization of Cloud Computing in Big Data Processing

487

In the formula,

represents the number of

decision trees in the framework. In the cloud

computing environment, the number of adjustment

trees can ensure the optimal performance of the

framework, and at the same time, control the

consumption of computing resources. In practice,

increasing the number of trees can improve the

accuracy of the framework, but it will consume more

cloud computing resources, such as when processing

user purchase predictions, if the number of trees is too

small, certain behavioral patterns will be missed.

()

Ly,FX;N Represents the loss function,

which is used to measure the probability of user

purchase and the error of actual purchase predicted by

the framework. Based on the minimization loss

function, the prediction accuracy of the framework

can be improved. In this way, in the cloud computing

environment, e-commerce platforms can.

Balance compute resources with the predictive

accuracy of the framework and provide accurate

purchase recommendations without sacrificing

responsiveness.

In the optimization, it is necessary to complete the

above content of tree depth optimization and

complexity control, for which Eq. (5) can be seen.

d arg min L(y, F(X;d))=

(5)

In the formula,

refers to the depth of the tree.

In cloud computing big data processing, the depth of

the tree controls the complexity of the framework. A

tree that is too deep will bring an overfit, and a tree

that is too shallow will not be able to capture enough

features. By adjusting the depth of the tree, the

framework will be able to capture key information

and reduce its computational resource usage. Based

on the depth of the optimization tree, the e-commerce

user behavior analysis platform can effectively

balance the complexity of the framework and

computing resources, improve the accuracy of its

prediction, and avoid the computational bottleneck

when processing large-scale e-commerce data.

Frame compression and parameter optimization

are also very important steps. In the optimization

process, by performing this step, the e-commerce

platform will effectively process the behavior data of

hundreds of millions of users in the cloud computing

environment, and significantly reduce storage and

computing costs without affecting the prediction

accuracy. See Eq. (6) for this.

compressed T i

Mmin||T||



)

In the formula,

compressed

is the compressed

frame. In the practical application of big data

processing, the compression framework can

significantly reduce the computing and storage

overhead. For example, in the recommendation

platform of an e-commerce platform, a compressed

framework can reduce storage requirements and

speed up response times, further improving the user

experience.

Indicates the number of parameters

in the decision tree. Based on pruning or other

compression techniques,

is the parameters of the

decision tree will be effectively reduced, and the

frame size will be reduced and the load of cloud

computing will be reduced. In practice, this means

that the framework can process user requests faster

and improve the platform's real-time responsiveness.

3.4 Integration of E-Commerce User

Behavior Analysis Platforms

The integration of this platform is mainly to define

the block interface. First, the data interface and

communication protocol between each block are

determined to ensure the smooth transmission of

information, such as the rapid interaction between the

request processing block and the data storage block,

and then effectively ensure the efficient acquisition

and processing of data. Based on this, the block

integration test is carried out, and all blocks are

integrated into the test environment to verify the

compatibility and data circulation of the interface.

Based on simulated user requests, the stability of the

platform can also be tested. In addition, in order to

further optimize the performance of the platform, the

performance of the platform should be optimized

according to the test results after the integration test,

specifically, it is necessary to improve the response

speed of the data analysis algorithm and

recommendation engine, and optimize the resource

allocation to better reduce latency and improve the

overall efficiency.

INCOFT 2025 - International Conference on Futuristic Technology

488

4 RESULTS AND DISCUSSION

4.1 The Case Results of Data

Processing

The object of this application case is a large e-

commerce enterprise M, whose main goal is to

optimize its personalized recommendation platform

and increase its overall sales based on the analysis of

user behavior, product sales, and regional visit

conversion rate. The user data of e-commerce

enterprise M is large-scale, covering multiple regions

and product categories, and at the same time, the

user's operations on the enterprise platform have

diversified characteristics, such as browsing, adding

shopping carts, and finally purchasing. Based on data

analysis, e-commerce company M hopes to improve

the purchase conversion rate of users and make more

accurate recommendations for different product

categories. Using the e-commerce user behavior

analysis platform built this time, e-commerce

enterprise M will realize the application and

optimization effect of cloud computing in big data

processing on its own website. From 2020 to 2023,

the e-commerce company M has developed

extremely well, and its number of users continues to

grow, as shown in Figure 1.

Data fluctuations.

The processing of cloud computing.

01234567 8

Figure 1: The development of the number of users of e-

commerce company M from 2020 to 2023

4.2 Data Analysis Such as Basic User

Operation Behavior

According to the case study, the dwell time of users

is closely related to their actions. For example, based

on user actions, the average dwell time of users who

make a purchase is 10 minutes, while the dwell time

of users who only click is 5 minutes. This indicates

an increase in dwell time, representing a more likely

user to complete a purchase, as shown in Table 1.

Table 1: User Operation Behavior and Dwell Time

The type of

operation

Average

dwell

time

(minutes)

Percentage of users (%)

Click Products 5 50

Add a

sho

cart

7 30

Complete your

urchase

10 20

Table 1 shows the analysis of users' operation

behaviors on the platform, such as clicking, adding to

shopping cart, and purchasing behaviors, the average

dwell time of each operation, and the corresponding

user proportion. Based on the table, it can be seen that

user operation behavior and dwell time are the key

contents of cloud computing in big data processing.

Click on the product, add the shopping cart, and

complete the purchase to promote the user's

consumption, as shown in Figure 2.

Figure 2: The promotion effect of clicking on the product,

adding the shopping cart, and completing the purchase on

the user's consumption

In this case, the sales of commodity categories can

also have an impact on cloud computing big data

processing. Specifically, electronics had the highest

sales of 5,000 yuan, while books had the lowest

average cart value of 35 yuan. This suggests that

while books products have a lower per-transaction

value, high-priced products contribute more to total

sales. The specific sales of each product category and

the average cart value are shown in Table 2.

Table 2: Sales of product categories

Product category Total

Sales

(

RMB

)

Average cart value (RMB)

Electronics 5000 200

Books 1500 35

clothing 3000 75

Application and Optimization of Cloud Computing in Big Data Processing

489

Table 2 shows the sales data of product categories,

such as the total sales of each type of product and the

average value of the shopping cart, which provides a

basis for optimizing product recommendations.

Based on the above data, the e-commerce user

behavior analysis platform will better understand the

user's preference for different product categories, and

provide support for subsequent product

recommendations.

4.3 Visits and Conversion Rate

Analysis by Region

From the analysis of visits and conversion rates in

each region, the northern region has the highest

conversion rate at 5.5%, although the southern region

has a higher number of visits, but the conversion rate

is lower. In addition, the southern region received

4,500 visits, but the conversion rate was the lowest at

3.2%, reflecting that users in the southern region are

more willing to compare prices and consider other

factors in the shopping process, as shown in Table 3.

Table 3: Visits and conversion rates by region

According to

the type.

Visits Conversion rate (%)

Optimize your

hone.

3000 5.5

Actively

optimize data.

4500 3.2

New phone. 4000 4.8

Table 2 shows the comparison of visits and

conversion rates in different regions, with the aim of

understanding user behavior patterns and purchasing

decisions in each region to adjust regional marketing

strategies. The relationship between visits and

conversion rate and purchase decisions is shown in

Figure 3.

Figure 3: The relationship between a user's visits,

conversion rate, and purchase decision

It can be seen that the conversion rate of users in

the northern region is the strongest, and the

conversion rate of users in the southern region is the

weakest. This indicates that the purchase leads in the

southern region need to be further optimized to

improve their overall conversion rate. Based on the

above data analysis, it can be concluded that user

dwell time has a significant impact on purchase

behavior, and the high unit price products in the

product category contribute significantly to sales. In

addition, regional marketing strategies need to be

adjusted according to their conversion rates to

improve overall sales performance. From the data, it

can be fully proved that this study is valid.

5 CONCLUSIONS

Based on this paper, the advantages of cloud

computing in big data processing have been fully

verified, especially in the analysis of user behavior on

e-commerce platforms. Cloud computing technology

makes distributed storage and computing possible,

effectively improving its data processing speed and

resource utilization. Moreover, according to the

research in this paper, based on the effective

combination of parallel computing and intelligent

algorithms, the platform can quickly analyze massive

user behavior data and optimize the recommendation

platform in real time. In the process of big data

processing, the elastic scheduling of resources and the

application of optimization algorithms will greatly

improve the response speed and accuracy of the

platform, and then provide high stability, high

flexibility and effective technical support for the

platform. In short, through the research of this paper,

it can be found that the application based on cloud

computing can not only help the platform achieve the

efficiency of big data processing, but also bring

higher sales performance and market competitiveness

to the platform. The research in this paper contains a

lot of data, but its cases are still limited, which makes

it inevitably limited, and it can be further studied in

the future to achieve effective optimization and

scaling applications.

ACKNOWLEDGEMENTS

2022 Guangzhou Higher Education Teaching Quality

and Teaching Reform Project - Department of

Computer Basics (Project No.: 2022KCJYS017);

2022 Guangdong Higher Education Teaching Reform

Project - Design and Implementation of the New

Talent Training Program Management System Based

on the OBE Concept (Project No.: C2206001163)

INCOFT 2025 - International Conference on Futuristic Technology

490

REFERENCES

Ahmad, A., R. A. Khan, S. U. Khan, H. S. Alwageed, A. A.

Al-Atawi, and Y. M. Lee. (2023). Green cloud

computing adoption challenges and practices. a client's

perspective-based empirical investigation. Cognition

Technology & Work, 25(4), 427-446.

Alshareef, H. N. (2023). Current Development, Challenges

and Future Trends in Cloud Computing. A Survey.

International Journal of Advanced Computer Science

and Applications, 14(3), 329-338.

Arulmozhiselvan, L., and E. Uma. (2024). Cloud and Fog

Computing Amalgamation for Data Agitation and

Guard Intensification in Health Care Applications. Ksii

Transactions on Internet and Information Systems,

18(3), 685-703.

He, Y. J., W. H. Ouyang, S. L. Li, L. Wang, J. Zhou, W. W.

Su, S. Z. Li, D. H. Mei, Y. Shi, Y. X. Jin, C. L. Li, and

Y. H. Ren. (2023). Cloud computing data privacy

protection method based on blockchain. International

Journal of Grid and Utility Computing, 14(5), 480-492.

Kang, H. Y., and J. Deng. (2023). A cross encryption

scheme for data security storage in cloud computing

environment. International Journal of Internet Protocol

Technology, 16(1), 1-10.

Kumar, K., and D. K. Saini. (2024). Security issues in

cloud, fog and edge computing models and suggested

solutions. Journal of Discrete Mathematical Sciences &

Cryptography, 27(4), 1247-1257.

Muhic, M., L. Bengtsson, and J. Holmström. (2023).

Barriers to continuance use of cloud computing.

Evidence from two case studies. Information &

Management, 60(5).

Nagahawatta, R., M. Warren, S. Salzman, and S. Lokuge.

(2024). Towards an Understanding of Cloud

Computing Adoption in SMEs. The Role of Security

and Privacy Factors. International Journal of Cyber

Warfare and Terrorism, 14(1).

Pericherla, S. S. (2023). Cloud Computing Threats,

Vulnerabilities and Countermeasures. A State-of-the-

Art. Isecure-Isc International Journal of Information

Security, 15(1), 1-58.

Pham, X. Q., T. Huynh-The, D. S. Kim, T. D. Nguyen, and

E. N. Huh. (2023). Distributed Cloud Computing.

Architecture, Enabling Technologies, and Open

Challenges. Ieee Consumer Electronics Magazine,

12(3), 98-106.

Application and Optimization of Cloud Computing in Big Data Processing

491