---

Santander Customer Transaction Prediction

Overview

This project aims to predict whether a customer will make a specific transaction using a binary classification model. The work includes data understanding, exploratory analysis, model development, and performance evaluation.

Summary

You can find the project summary in 02_Non-Technical-Summary.md

PRD

All the product requirements are in 03_PRD.md

Requirements

• Python 3.11.3
• All dependencies are listed in requirements.txt

Setup Instructions

1. Create and Activate a Virtual Environment

Before running the notebooks, fork the repository and set up a fresh virtual environment.

2. Install Dependencies

Use the commands below for your operating system.
If installation errors occur (especially on Apple Silicon), removing strict version pins in requirements.txt may help.

---

Here's a combined README with all necessary information organized clearly:

---

Environment Setup

Please make sure you have forked the repo and set up a new virtual environment.

Note:

• If there are errors during environment setup, try removing the versions from the failing packages in the requirements file.
• In some cases it is necessary to install the graphviz compiler for the transformers library.
• Make sure to install hdf5 if you haven't done it before.

Prerequisites: Install Graphviz and HDF5

Check if graphviz is already installed by running:

dot -V

If you haven't installed it yet, follow the instructions below for your operating system.

macOS

Update Homebrew and install graphviz and hdf5:

brew update
brew install graphviz
brew install hdf5

Restart your terminal and verify the installation:

dot -V

Windows

Update chocolatey and install graphviz:

choco upgrade chocolatey
choco install graphviz

Press Y for the standard installation.

For hdf5, visit this website to install hdf5.

Restart your terminal and verify the installation:

dot -V

Linux

Ubuntu / Debian

sudo apt-get update
sudo apt-get install -y build-essential cmake libomp-dev graphviz libhdf5-dev

Fedora / RHEL / CentOS

sudo dnf install -y gcc-c++ cmake libomp graphviz hdf5-devel

Python Environment Setup

macOS

For macOS with Intel chips:

pyenv local 3.11.3
python -m venv .venv
source .venv/bin/activate
pip install --upgrade pip
pip install -r requirements.txt

For macOS with Silicon chips (M1/M2/M3):

pyenv local 3.11.3
python -m venv .venv
source .venv/bin/activate
pip install --upgrade pip
pip install -r requirements_silicon.txt

If LightGBM fails to install:

brew install cmake libomp
pip install lightgbm
pip install -r requirements.txt

Windows

PowerShell:

pyenv local 3.11.3
python -m venv .venv
.venv\Scripts\Activate.ps1
python -m pip install --upgrade pip
pip install -r requirements.txt

Git Bash:

pyenv local 3.11.3
python -m venv .venv
source .venv/Scripts/activate
python -m pip install --upgrade pip
pip install -r requirements.txt

If LightGBM fails due to CMake or compiler errors:

pip install cmake
pip install lightgbm

Linux

pyenv local 3.11.3
python -m venv .venv
source .venv/bin/activate
pip install --upgrade pip
pip install -r requirements.txt

If LightGBM fails to install:

pip install lightgbm

---

Models Evaluated

• Logistic Regression — baseline linear model.
• LightGBM — gradient-boosted tree model optimized for speed and efficiency, designed to handle large datasets and high-dimensional features with fast training and strong predictive performance.
• XGBoost — tree-based boosting model used to benchmark performance and validate robustness across model families.

---