Tuijin Jishu/Journal of Propulsion Technology

ISSN: 1001-4055
Vol. 45 No. 2 (2024)
___________________________________________________________________

Amazon User Segmentation

Ankit Chaudhary#1, Abhishek Pal#2, Ankit Saraswat#3, Harshit Jindal#4, Jagbeer Singh#5
#
Computer Science and Engineering, Meerut Institute of Engineering and Technology (MIET),
Meerut, India

Abstract— In e-commerce environments, user experiences and marketing strategies are best optimized by
effective user segmentation. This project classifies users according to pertinent criteria like age and buy rating by
using sophisticated clustering algorithms like K-means and DBSCAN. The Elbow Method makes it easier to
determine the ideal number of clusters, resulting in a more accurate and detailed segmentation. Disk plots and
other visualizations provide information about different user groups and their purchase patterns. Interactive user
interfaces make these segments easier to explore.
The effectiveness of K-means is compared with hierarchical clustering and DBSCAN in a thorough analysis that
includes criteria such as the Silhouette Score for reliable cluster evaluation. We go into great detail about ethical
issues, such as algorithmic fairness and user privacy. By strengthening our knowledge of user behavior in the
context of e-commerce, this study lays the groundwork for customized marketing campaigns that cater to
individual user preferences.

Keywords— E-commerce, DBSCAN, User segmentation, silhouette metric, clustering technique, Amazon

I. Introduction
In the dynamic landscape of e-commerce, understanding user behavior and preferences is crucial for delivering
personalized experiences. This project delves into the segmentation of Amazon users, employing advanced
clustering techniques to categorize users based on age, income, and the purchase rating.

1.1. Background History

User segmentation has evolved as a strategic approach for businesses to tailor their offerings and communication.
The Amazon, being a pioneer in e-commerce, constantly seeks innovative methods to enhance user experience.
This project aligns with the broader trend of employing the machine learning for a precise user targeting.

Supported Technologies and Algorithms

The project harnesses the power of Python for its versatility and rich ecosystem of libraries. Key libraries such as
NumPy and pandas facilitate data manipulation, while Matplotlib and Plotly Express enable insightful
visualizations. Scikit-learn provides robust implementations of clustering algorithms like K-means and DBSCAN.

II. Literature Review

In e-commerce, user segmentation has undergone a radical change, moving from conventional demographic-
centric techniques to advanced machine learning methodologies. As e-commerce platforms grew, it became clear
that the early tactics' limitations—which mostly relied on characteristics like age and income—were
unsatisfactory. This led to the investigation of more complex and dynamic segmentation techniques. Automating
user classification based on shared attributes has been made possible by machine learning, especially with regard
to clustering methods.
K-means is one of the most popular clustering algorithms since it's easy to use and works well with large datasets.
But its sensitivity to beginning centroids and presumptions of spherical and uniformly sized clusters led to the

1279
Tuijin Jishu/Journal of Propulsion Technology
ISSN: 1001-4055
Vol. 45 No. 2 (2024)
___________________________________________________________________

investigation of substitute techniques. A noteworthy answer was offered by density-based clustering, which was
demonstrated by DBSCAN. This method identified clusters based on data density. This methodology not only
exhibited resilience to anomalies but also exhibited adaptability in identifying groups of any kind, conforming to
the varied and complex user conduct patterns observed in actual situations.
Current developments in user segmentation involve merging data from several sources, including past purchases,
preferences, and user interactions. Neural network-based clustering is one example of a deep learning technique
that has been studied for its ability to find complex correlations in user data, perhaps leading to improved
segmentation accuracy.
But difficulties also accompany advancement. One of the key concerns that still exists is the interpretability of
clusters, which is essential for practical insights. More complexity in this area is related to handling high-
dimensional data, maintaining privacy compliance, and meeting the need for real-time segmentation to adjust to
changing user behavior.
The literature study concludes by outlining the development of user segmentation techniques in e-commerce and
highlighting the field's dynamic character. Though K-means clustering is still a popular option, research into
density-based techniques such as DBSCAN, along with developments in deep learning and integrating several
data sources, suggests a search for more flexible and precise segmentation methods. Obstacles and unanswered
concerns highlight the necessity of ongoing investigation and creativity in user segmentation approaches, offering
a solid basis for next studies targeted at improving tailored advertising and user experience on e-commerce sites.

III. Related Work

Extensive time has been devoted by researchers to the in-depth study of market segmentation. The idea of
marketing segments was initially introduced by Smith (1956) as a mechanism to underscore the diversity and
distinctions in consumer desires. It was acknowledged that this approach facilitates the satisfaction of a wide array
of consumer needs. However early studies on segmentation did not thoroughly investigate the factors that
contribute to it. Segmentation variables encompass a selected set of characteristics such, as usage patterns,
psychographics, demographics and lifestyles. These characteristics help classify units into groups ensuring that
each group consists of units That are most similar, to each other (Chan et al., 2012).Several effective segmentation
strategies have emerged, drawing on segmentation variables as outlined by Kotler and Armstrong in 1999, and
Cahill in 2006. Becker and Conner, in 1981, employed "personality variables" for consumer market segmentation,
whereas Anderson et al. in 1993 categorized users through financial and unique segmentation var. Based on Cahill
(2006), segmentation themes that meet specific criteria can be broadly categories into 2 groups: those centered
around user behavior, such as life-style, stages of life, psychographics, and patterns of use, and those determined
by physical characteristics, such as demography, location, or a combination of the two. Nevertheless, commonly
utilized factors in published works often fall short in distinguishing between user psychographic categories,
rendering them ineffective in accurately identifying submarkets (Kotler and Armstrong, 1999). Within the
framework of market segmentation theory, segmentation variables can be seen in several empirical investigations
in addition to conceptual studies. The segmentation variables that are most commonly used in empirical
investigations are psychographics, values/attitudes, and demographic, regional, and economic aspects (Hassan
and Craft 2005). (Abratt 1993). Understanding the needs of user groups is crucial when creating marketing plans.
Market segmentation studies help us achieve this by allowing us to target segments, with precision ultimately
increasing user value (Hwang et al., 2004). Several market models have been developed to help in formulating
marketing strategies for segments. For example, The RFM model leverages variables, whereas the LTV model
concentrates on value-related variables. However previous research has primarily focused on market segmentation
in the B2B context than exploring its application to product introductions (Kumar and Reinartz 2012). Businesses
can effectively reach groups of users by employing segmentation methods (Foedermayr and Diamantopoulos,
2008). The specific objectives of the study will determine the choice of segmentation techniques to be used
(Tsiptsis, and Chorianopoulos, 2010). For needs-based and attitudinal segmentation are frequently utilized within
companies to support initiatives. Conversely, behavioral segmentation serves as an approach used to formulate

1280
Tuijin Jishu/Journal of Propulsion Technology
ISSN: 1001-4055
Vol. 45 No. 2 (2024)
___________________________________________________________________

customized strategies for personalized product offerings. The ITF model has been discovered to be applicable, in
identifying user innovation segments. in online communities in a study conducted by Chen et al. (2018).

Table 1. A comprehensive overview of the existing literature, on market segmentation, throughout time
periods.
IV. Proposed Work Plan

A. GENERAL ARCHITECTURE,
The overall design consists of a methodical flow of operations. Preprocessing the data sets the stage for the Elbow
Method's application, which yields the ideal cluster numbers. After that, K-means clustering is carried out, and
Plotly Express is used to show the outcomes. One way to quantify the quantitative quality of a cluster is through
the Silhouette Score.

B. DESCRIPTION OF VARIOUS MODULES

● Data Processing
Important customer variables including age, wealth, and buy rating are analyzed using data from Amazon. In
order to guarantee the dataset's integrity and preparation for further analysis, the first step of the process entails
careful data processing that includes thorough cleaning and categorization.

● Elbow Method Analysis

The Elbow Method is a crucial analytical tool that guides the process of figuring out how many clusters the K-
means algorithm should have. This method lays the foundation for successful user segmentation by carefully
examining a range of cluster counts, striking a subtle balance between simplicity and precision.

● K- Means Clustering

1281
Tuijin Jishu/Journal of Propulsion Technology
ISSN: 1001-4055
Vol. 45 No. 2 (2024)
___________________________________________________________________

The K-means clustering technique is the foundation of user segmentation. Based on purchase rate and age, this
approach, which is based on centroid-based clustering, divides users into groups. Cluster assignments are refined
iteratively to guarantee a complete segmentation procedure that culminates in discrete user segments.

● Visualization

Plotly Express is skillfully used to create dynamic visual representation that is visually striking. Interactive
components give visually striking scatter plots that bring user clusters to life. This gives customers the ability to
interactively examine and evaluate individual data points in addition to giving them a perceptive overview of the
segmentation results.

● Silhouette Score Calculation,

The Silhouette Score is a crucial numerical indicator that measures how effective the clustering procedure was.
This score provides a nuanced indication of the isolation and distinctiveness of the clusters that have been detected.
Examining the Silhouette Score provides the project with important information about the overall effectiveness
and coherence of the user segmentation that was accomplished.

C. ALGORITHM OF MAIN COMPONENT

The K-means clustering technique is the central component of the project. While the Silhouette Score measures
the quality of the clustering, utilizing the Elbow Method guarantees a well-informed choice of cluster count.

V. Experimental Test Analysis

A. DESCRIPTION OF DATA SET USED
The dataset used in this study is a comprehensive collection of user data that was obtained from Amazon.com.
Important characteristics consist of:
● Age: This indicates the range of demographics among the Amazon customers.

1282
Tuijin Jishu/Journal of Propulsion Technology
ISSN: 1001-4055
Vol. 45 No. 2 (2024)
___________________________________________________________________

● Income: It offers information on the financial backgrounds of users.

● Purchase rating: It provides a numerical indicator of users' pleasure and involvement.
In order to guarantee a representative sample and allow for thorough analysis and user segmentation, this dataset
has been carefully chosen.

B. SYSTEM EFFICIENCY EVALUATION

The Silhouette Score, an important metric for evaluating the caliber of clusters created by the clustering
algorithms, is used to analyze the effectiveness of the system that was constructed. The cohesiveness within
clusters as well as the distance between clusters are both considered by the Silhouette Score. A higher Silhouette
Score denotes clear and well-defined clusters, indicating how well the user segmentation worked.
Each data point in the procedure has its Silhouette Score calculated, and the average score is a general indicator
of the quality of the grouping. This assessment measure helps determine the ideal number of clusters and assesses
how effectively the selected attributes support the creation of significant user segments. The efficiency analysis
provides a numerical framework for evaluating how well the clustering algorithms identify patterns in the dataset.
The K-means algorithm involves mathematical formulas for updating cluster centroids and assigning data points
to clusters.
The flow chart of the project is:

1283
Tuijin Jishu/Journal of Propulsion Technology
ISSN: 1001-4055
Vol. 45 No. 2 (2024)
___________________________________________________________________

1. Euclidean Distance:
The Euclidean distance between two points P(x1,y1) and Q(x2,y2) in a two-dimensional space is given by:
Distance = √(𝑥2 − 𝑥1 )2 + (𝑦2 − 𝑦1 )2
• This distance measure is used to determine the proximity of data points to cluster centroids.

2. Cluster Centroid Update:

For each cluster k, the centroid is updated as the mean of all data points assigned to that cluster:
1
Centroidk = 𝛴𝑝𝑜𝑖𝑛𝑡𝑠 𝑖𝑛 𝑐𝑙𝑢𝑠𝑡𝑒𝑟𝑘
𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑝𝑜𝑖𝑛𝑡𝑠 𝑖𝑛 𝑐𝑙𝑢𝑠𝑡𝑒𝑟𝑘

3. Assigning Data Points to Clusters:

Each data point is assigned to the cluster whose centroid is closest. The assignment is based on minimizing the
Euclidean distance between the data point and each cluster centroid.

4. Within-Cluster Sum of Squares (WCSS):

WCSS is used in the Elbow Method to find the optimal number of clusters. It is the sum of the squared distances
between each data point in a cluster and its centroid:

WCSS = ∑𝑝𝑜𝑖𝑛𝑡𝑠 𝑖𝑛 𝑎𝑙𝑙 𝑐𝑙𝑢𝑠𝑡𝑒𝑟 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 𝟐

4. Silhouette Score:
𝑏(𝑖)−𝑎(𝑖)
𝒔(𝒊) =
max (𝑎(𝑖),𝑏(𝑖))
where:
• s(i) is the silhouette score for sample i.
• a(i) is the average distance from sample i to other samples in the same cluster.
• b(i) is the smallest average distance from sample i to samples in a different
cluster, minimized over clusters.
VI. Conclusion

Research uses advanced clustering techniques to give a comprehensive approach to user segmentation on
Amazon.com. Important insights into the preferences and behavior of users are obtained by means of the
Silhouette Score and the outcome’s analysis. This establishes the framework for upcoming improvements and
practical uses of the concept. This research not only helped us gain a better understanding of how users behave
on Amazon.com, but it also set the stage for more individualized marketing approaches and enhanced user
interfaces. Prospective directions for future research encompass investigating other clustering techniques,
integrating real-time data for dynamic segmentation, and working with industry partners for pragmatic
application. Machine learning applied to customer segmentation is a modern, data-driven strategy in e-commerce
that promises more developments and insights for improving user experiences and happiness. Encouragingly, the
User Segmentation Project represents a major advancement in the optimization of user-centric methods within the
ever-changing e-commerce ecosystem.

1284
Tuijin Jishu/Journal of Propulsion Technology
ISSN: 1001-4055
Vol. 45 No. 2 (2024)
___________________________________________________________________

References
[1] 1956, R. Smith explored the theme of " Product Distinctiveness and Market Subsegmentation as Alternative
Marketing Strategies" in the Journal of Marketing (vol. 21, pp. 3-8).
[2] Kotler, and G. Armstrong, Principles of Marketing: Upper Saddle River . NJ: Prentice Hall, 1999.
[3] J. Cahill, Lifestyle market segmentation. New York: Haworth Press, 2006.
[4] Anderson, D. C. Jain, and P. K. Chintagunta, "Customer value assessment in business markets: a state-of-
practice study," Journal of Business-to-Business Marketing, vol. 1, pp. 3-29, 1993.
[5] S. Hassan, and S. H. Craft, "Linking global market segmentation decisions with strategic positioning options,"
Journal of Consumer Marketing, vol. 22, pp. 81-89, 2005.
[6] Abratt, "Market segmentation practices of industrial marketers," Industrial Marketing Management, vol. 22,
pp. 79-84, 1993.
[7] K. Foedermayr and A. Diamantopoulos, "Market segmentation in practice: review of empirical studies,
methodological assessment, and agenda for future research," Journal of Strategic Marketing, vol. 16, pp. 223-
265, 2008.
[8] Kumar, and W. Reinartz, "Customer relationship management issues in the business-to-business context,"
Customer Relationship Management, pp. 261-277, 2012.
[9] Tsiptsis, and A. Chorianopoulos, Data Mining Techniques in CRM: inside Customer Segmentation. Wiley
Publishing, 2010.

1285