🔬 BMLP Project Submission

🧠 Project Overview

This project is a submission assignment. This project analyzes customer behavior by grouping several customers into several criteria groups. This project uses the clustering method using K-Means Clustering and then analyzed using Classification.

🎯 Objectives

• Group data using K-Means Clustering
• Analysis using several Classification models

📁 Dataset

Bank Transaction Data

🛠️ Methodology

• Exploratory Data Analysis (EDA)
  • Review the number of rows columns and data types in the dataset
  • Display descriptive statistics of the dataset
  • Display correlations between features
  • Display histograms for all numeric columns
  • Visualize correlation matrix using heatmap
• Data Preprocessing
  • Handling missing values
  • Handling duplicate data
  • Handling outliers
  • Feature scaling using Standard Scaler
  • Drop in the column that contains the ID and IP Address information
  • Perform feature encoding using Label Encoder for categorical features
• Clustering Model Development
  • Feature Scaling is done first, then the Elbow Method visualization is performed
  • Clustering using the K-Means Clustering algorithm
  • Creating a visualization of clustering results
• Interpretation of clustering results
• Inverse dataset to normal range for numeric and inverse encoded dataset to its original category
• Reinterpretation after inverse
• Classification Model Development
  • Data Splitting
  • Create a classification model using Decision Tree
  • Training models using classification algorithms other than Decision Tree
  • Displays the results of evaluating accuracy, precision, recall, and F1-Score
  • Saving models using joblib
  • Perform Hyperparameter Tuning
  • Saving Hyperparameter Tuning using joblib

📈 Results

• Cluster 0 Analysis:
  • Interpretation: High-Income Cluster
  • This cluster has a moderate number of transactions, not more than cluster 1 or a little like cluster 2. Likewise, in CustomerAge which has an adult age range that is not too old and a little at a young age. This cluster is very rich and has a lot of savings, unlike cluster 1 which has little savings or cluster 2 which has savings that are fairly moderate.Then this cluster is dominated by the professions of doctors and engineers who have high incomes in the CustomerOccupation feature.
• Cluster 1 Analysis:
  • Interpretation: High-Spending Youth Cluster
  • This cluster has characteristics with a fairly large number of transactions. It is considered very wasteful because the number of transactions is not balanced with the amount of savings owned. It can be seen in the TransactionAmount feature with the highest data distribution but the AccountBalance feature has the lowest data distribution. This cluster has an age range of only young people in the CustomerAge feature. This cluster is dominated by students and college students in the CustomerOccupation feature who basically do not have an income.
• Cluster 2 Analysis:
  • Interpretation: Retiree Cluster
  • This cluster is the most economical cluster with a small number of transactions and moderate savings. Seen in the TransactionAmount feature with the lowest data distribution and in the AccountBalance feature has a medium data distribution. This cluster has an old adult age range in the CustomerAge feature. This cluster is dominated by retired professions who are no longer working and followed by the engineering profession.