MY DETAILS
Name: Niveethita A/P Pandia Rajan
Matric Number: A24CS0148
Course: Computer Science (Data Engineering)
Programme: Data Mining
Section: 02
BRIEF INTRODUCTION
Hello! I am Niveethita, currently exploring the field of data engineering by working on projects involving database design, data processing and data analytics.
Through the Data Mining course, I was able to solidify my understanding of data processing and gained hands-on experience with the complete data pipeline from preparing raw data to performing data analysis.

Dirty Dataset Challenge
From working on the employee salary database, I learned how to identify and fix real-world data quality issues such as missing employee names, duplicate employee records, and inconsistent data.
I applied preprocessing techniques like filling missing salaries using the median salary of each department and removing duplicate records to make the dataset clean and reliable.
Through this project, I realized that every data cleaning step affects the quality of the analysis and accurate data is essential for producing quality insights and supporting better business decisions.
Classification Class Activity
Through the classification activity, I learned how supervised learning algorithms such as Decision Trees and Naïve Bayes use historical data to classify new data into predefined classes.
I also learned the process of training these models, including selecting relevant features, applying cross-validation, and evaluating their performance to build more reliable models.
Through this activity, I realized the importance of evaluation metrics such as the confusion matrix, precision, and recall, as they help measure how well a model performs, especially when dealing with imbalanced datasets or situations where false positives and false negatives can have significant effects.
ASSIGNMENT 1
Assignment 1
Through this assignment, I gained experience in analyzing large datasets, including e-commerce consumer behavior and social media user profiles, to identify useful patterns and trends. I learned how proper data preprocessing and selecting relevant attributes play an important role in producing accurate analysis results.
This activity also helped me understand how data mining can be applied in real-world situations, where insights from customer behavior can be used to improve marketing strategies, enhance customer engagement, and support better business decisions.
ASSIGNMENT 2 & TEST
Assignment 2
Through this assignment, I gained a better understanding of how association rule mining can be used to discover relationships between items in transaction datasets.
I learned how algorithms such as FP-Growth and ECLAT work differently to identify frequent patterns, where FP-Growth uses a compact tree structure to reduce candidate generation, while ECLAT applies a vertical database approach using Transaction ID (TID) set intersections.
This activity also helped me understand that choosing the right algorithm depends on the characteristics of the dataset, such as its size and density, to achieve more efficient pattern mining.
Test
The midterm test covered Modules 1 to 4, including data mining fundamentals, data preprocessing, classification, and clustering. Preparing for the test helped me strengthen my understanding of these topics and it was a good way to review and recall the concepts that were discussed in class. The final examination covering Modules 2 to 5 was again solidifying the concepts and helped to deepen my understanding on the data mining pipeline.
PROJECT 1
Loan Approval Prediction (Supervised Learning)
This project focused on predicting whether a loan application would be approved using machine learning classification techniques. The dataset used was the Loan Prediction Dataset, which contains 614 applicant records with information related to demographics, income, loan details, education, and credit history.
The data was first preprocessed by handling missing values using mode and median imputation, followed by feature engineering to create new attributes such as TotalIncome and LoanIncomeRatio.
Six classification models were developed and compared, including Logistic Regression, Decision Tree, Random Forest, Gradient Boosting, SVM, and KNN, with performance evaluated using Stratified 5-Fold Cross-Validation. The results showed that SVM achieved the highest overall accuracy and F1-score, while Random Forest performed best in precision and ROC-AUC.
From this project, I learned that credit history plays a major role in loan approval prediction and that model selection depends on balancing prediction performance with the risks of false approvals or rejections.
PROJECT 2
Online Retail Customer Segmentation (Unsupervised Learning)
For the second project, my group explored a different domain by using the Online Retail Dataset to gain experience with a new type of dataset and uncover different business insights.
The dataset contains transactional information such as customer ID, quantity purchased, unit price, and invoice date from an e-commerce platform. Data preprocessing was performed by removing incomplete records, duplicate transactions, and invalid values, followed by creating RFM (Recency, Frequency, Monetary) features to represent customer purchasing behavior.
K-Means Clustering and Agglomerative Hierarchical Clustering were applied to group customers into three segments, with their performance compared using clustering evaluation metrics. K-Means produced better clustering results with a higher Silhouette Score and lower Davies-Bouldin Index.
The analysis identified three customer groups: Premium Customers who contribute high revenue, Standard Customers who provide consistent sales, and At-Risk Customers who require re-engagement strategies.
By working with a different application domain, I gained a broader understanding of how clustering techniques can be used to generate valuable business insights and support data-driven marketing and customer retention strategies.
INDIVIDUAL CONTRIBUTION
For Assignment 1, my group compared two different datasets. I was responsible for analyzing the Social Media User Analysis dataset, while my teammate worked on the E-commerce Consumer Behaviour dataset. I completed the explanation and analysis for my assigned dataset, and both of us worked together on the discussion and conclusion by comparing the insights from the two datasets.
For Assignment 2, I was responsible for explaining the ECLAT algorithm, including how it works and its advantages. I also integrated all of my group members' content into the final poster, ensuring the information was well-organized, concise, and easy for viewers to understand.
For Project 1 (Loan Approval Prediction), I was responsible for the model performance evaluation. I evaluated the classification models using metrics such as Accuracy, Precision, Recall, F1-Score, and ROC-AUC, analyzed the results, and provided suggestions to improve the models so that they better met the project requirements and assessment rubric.
For Project 2 (Online Retail Customer Segmentation), I was involved in developing both the K-Means and Agglomerative Hierarchical Clustering models and evaluating their performance using the Silhouette Score and Davies-Bouldin Index. I also helped organize the overall project workflow to ensure each stage of the data pipeline, from preprocessing and feature engineering to model evaluation, was completed in a consistent way.
REFLECTION
Throughout these projects, I gained a better understanding of the complete data mining pipeline, from data preprocessing and feature engineering to model development, evaluation, and interpreting the results. Working on real datasets gave me hands-on experience in applying data mining techniques to solve practical problems and showed me how meaningful insights can be extracted to support better business decisions.
One of the biggest challenges I faced was finding a suitable dataset that could provide valuable and meaningful insights. The data cleaning process was also difficult at the beginning, especially when dealing with missing values, duplicates, and inconsistent data. Although there was a steep learning curve, I was able to overcome these challenges with the help of my classmates, guidance from my lecturer, and additional learning from online resources.
For future improvement, I would like to explore more advanced data mining models and techniques to broaden my knowledge. I also hope to work with larger and more complex datasets so that I can continue improving my analytical skills and gain more experience in solving real-world data problems.