Economic Preferences Clustering Using Generative and Deep Learning Models

This repository contains the implementation of a research project that explores robust clustering techniques to identify economic preference types using generative and deep learning-based methods. The project uses a dataset of economic preferences and applies various clustering algorithms such as Gaussian Mixture Models (GMM), Wishart Mixture Models (WMM), and Variational Deep Embedding (VaDE) to uncover patterns in economic preferences.

---

Features

• Economic Preference Analysis: Analyze risk aversion, time preference, and social preference.
• Clustering Algorithms:
  • Gaussian Mixture Models (GMM) implemented using Expectation-Maximization (EM).
  • Wishart Mixture Models (WMM) implemented using Generalized Expectation-Maximization (GEM).
  • Variational Deep Embedding (VaDE) for generative deep clustering.

---

Files and Descriptions

1. main.py: Main source code including preprocessing, clustering, and visualization.
2. mm_em_gaussian.py: Implementation of Expectation-Maximization Gaussian Mixture Model (EM-GMM).
3. mm_gem_wishart.py: Implementation of Generalized Expectation-Maximization Wishart Mixture Model (GEM-WMM).
4. vade_main.py: Main script for training and evaluating VaDE.
5. vade_pretrain.py: Pretraining script for parameter initialization in VaDE.
6. vade_model.py: Implementation of the Variational Deep Embedding (VaDE) model based on the paper "Variational Deep Embedding: An Unsupervised and Generative Approach to Clustering". (Adapted from mori97/VaDE).
7. transformed_features.txt: Dataset of transformed features after preprocessing the original data (used in VaDE).

---

How to Use

Ensure you have Python 3.7.1 installed along with the following dependencies:

• matplotlib 3.5.2
• numpy 1.21.6
• scikit-learn 0.22.1
• scipy 1.7.3
• tensorboard 2.9.1
• tensorboardX 2.5.1
• torch 1.2.0+cu92
• torchvision 0.4.0+cu92