Analyzing and Presented By Supervised By
predicting consumer
behavior using ERUM EJAZ FATIMA
machine learning 2221265
Sir Tahir Abdullah
2022-GCUF-00627
Outline
• Introduction
• Literature Review
• Problem Statement
• Research Objective’s
• Methodology
• Framework
• Results
• Conclusion
• References
Introduction
• This research explores churn prediction using
machine learning to analyze a Finnish insurance
firm's customer attrition. It compares the firm's
current methods with literature-recommended
approaches and finds that random forests
outperform other models in predicting churn. (Gao
et al., 2022).
The study suggests that machine learning offers a
slightly better alternative for estimating client
turnover compared to existing methods
Problem Statement
The rise of subscription-based business
models due to digitalization has led to
increased competition and customer churn,
especially in B2B contexts.
While retaining customers is more cost-
effective than acquiring new ones, predicting
and managing churn remains challenging.
Literature Review
Author
Title Year Dataset Algorithms Results Future Work
Name
Walmart Random Forest The analysis Using more advanced
Walmart Data Md 2023
dataset Decision Tree revealed that machine learning
Analysis Using Humayun
Logistic holidays have a techniques, such as
Machine Learning Kabir
Regressor significant impact deep learning
Linear on Walmart's
Regressor sales, with a
Support Vector notable increase
Machines in sales during
(SVM) holiday periods.
Utilizing Machine Zhang online Supervised Improved Explainable AI (XAI)
2023
Learning Algorithms Yixuan transactio learning customer techniques to make
for Consumer ns, social models segmentation and machine learning models
Behavior Analysis media Unsupervised targeting more transparent and
posts, learning Enhanced interpretable
search models recommendation Federated learning to
Deep learning systems enable collaborative
queries,
models Better prediction training of machine
and e-
of customer churn learning models without
commerc and purchase sharing sensitive data
e behavior
platforms
historical Logistic accuracy of 0.85, authors suggest that the
Customer Dr. Deepti 2024
purchasin Regression precision of 0.84, developed models can be
Purchasing Sharma
g data Decision Tree recall of 0.83, and integrated into business
Behaviour
from Random Forest F1 score of 0.83. decision-making
Observation: Using
various K-means processes to enhance
Machine Learning
sources clustering marketing strategies and
Algorithms And
such as (mentioned in
Python customer satisfaction
transactio the literature
Implementation
n records, review)
Cont..
Title Author Name Year Dataset Algorithms Results Future Work
Predicting Customer Vidya 2023 e-commerce Logistic Regression, accuracy of 87.83% in future studies may
Churn in E- Rajasekaran dataset with Kaplan-Meier curve predicting customer examine real-time data
Commerce Using customer analysis churn analytics, sophisticated
Statistical and information deep learning
Machine Learning methods, and models
Methods that adjust to shifting
consumer behavior as
the e-commerce
industry changes.
Customer data Md Abdullah Al 2024 Customer Support Vector accuracy of 83.45% Exploring other
prediction and Rahib churn Machine, K-Nearest R2-score of 0.9782 machine learning
analysis in e- dataset Neighbour, Decision algorithms for
commerce using (5630 Tree, Logistic customer churn and
machine learning samples, 5 Regression, LSTM. product on-time
parameters) delivery prediction
Customer Investigating the
annual impact of additional
spending features on the
dataset performance of the
(448 models
samples, 4
parameters)
Product on-
time
delivery
dataset
(10999
samples, 7
parameters)
Research Objectives
• The present research aims to forecast future
customer status (stays/churns) or churns for
insurance customers for the upcoming year while
they are purchasing new private insurance, such as
property, life, or auto insurance.
• The model should be able to predict the churn for
both new and old customers. A thorough series of
machine learning algorithms has been applied on
the customer sales dataset to find the new features
that make the customer to churn.
• On the basis of this I discovered the most accurate
model that highlights some important churn factors
in e-commerce sector.
Materials and Methods
In supervised learning the characteristics of the data
are learned via the use of available labels and the
training process is used for predicting the importance
of another data point that has not yet been seen.
According to Sahar F. Sabbeh (2023) the following
supervised classification techniques are most often
employed to forecast consumer behavior:According to
Sahar F. Sabbeh (2023) the following supervised
classification techniques are most often employed to
forecast consumer behavior.
Framework
Cont..
• Dataset (Telco-Customer-Churn from Kaggle): The process
begins with the Telco-Customer-Churn dataset, which is a widely-
used collection of facial expressions.
• Data Preparation: This step involves cleaning and organizing the
data into a format suitable for analysis. It can include handling
missing values, data normalization, and data transformation.
• Feature Extraction: In this phase, relevant features or attributes
are selected or extracted from the data that is used for training the
ML model.
Cont..
• Splitting Data: The prepared and feature-extracted data
is then split into two sets: a training set and a testing set.
The training set is used to train the machine learning
model, while the testing set is used to evaluate the
model's performance.
• Machine Learning Techniques: Various machine learning
techniques have been applied to the training set to learn from the
data. These techniques include algorithms such as K-Nearest
Neighbors (KNN), Support Vector Machines (SVM), Logistic
Regression (LR), and Random Forest (RF).
• Performance Evaluation: Finally, the performance of the
machine learning model has been evaluated using the testing set.
Metrics such as accuracy, precision, recall, and F1-score were
used to assess how well the model performs classification tasks.
Results
Results
Conclusion
The confusion matrix shows 1400 correct non-churn
and 324 correct churn predictions, with 149 false
churn and 237 false non-churn predictions. To reduce
customer churn, firms should know their customers,
improve customer service, build loyalty, and survey
former customers to understand their reasons for
leaving.
References
• Alsagri, H., & Ykhlef, M. (2020). Machine Learning-based
Approach for Depression Detection in Twitter Using Content
and Activity Features.
https://www.who.int/en/news-room/fact-sheets/detail/depressi
on
• Faryal, M., Iqbal, M. M., & Tahreem, H. (2021). Mental Health
Diseases Analysis on Twitter using Machine Learning.
http://iksp.org/journals/index.php/ijcse/index
• Intelligence and Neuroscience, C. (2023). Psychological
Analysis for Depression Detection from Social Networking
Sites. Computational Intelligence and Neuroscience, 2023, 1–
1. https://doi.org/10.1155/2023/9796187
• Kumar, P., Samanta, P., Dutta, S., Chatterjee, M., & Sarkar, D.
(2022). Feature Based Depression Detection from Twitter
Data Using Machine Learning Techniques. Journal of Scientific
Research, 66(02), 220–228.
https://doi.org/10.37398/jsr.2022.660229
CONT..
• Obagbuwa, I. C., Danster, S., & Chibaya, O. C. (2023). Supervised
machine learning models for depression sentiment analysis.
Frontiers in Artificial Intelligence, 6.
https://doi.org/10.3389/frai.2023.1230649
• Singh, J., Wazid, M., Singh, D. P., & Pundir, S. (2022). An embedded
LSTM based scheme for depression detection and analysis.
Procedia Computer Science, 215, 166–175.
https://doi.org/10.1016/j.procs.2022.12.019
• Sofia, Malik, A., Shabaz, M., & Asenso, E. (2023). Machine learning
based model for detecting depression during Covid-19 crisis.
Scientific African, 20. https://doi.org/10.1016/j.sciaf.2023.e01716
THANK YOU!
