Interactive machine learning visualization dashboards built with Plotly Dash and scikit-learn.

Overview

A comprehensive collection of 4 interactive machine learning visualization applications developed for the “Interactive Data Visualization Analysis” course at University Leipzig, Germany. Each project demonstrates different ML algorithms through real-time interactive dashboards with parameter adjustment capabilities.

Technology Stack

  • Visualization: Plotly Dash - Interactive web applications
  • Machine Learning: scikit-learn - ML algorithms and evaluation
    • Linear Regression
    • K-Means Clustering
    • Logistic Regression
    • Decision Tree
    • K-Nearest Neighbors
    • Neural Network
  • Data Processing: pandas & NumPy - Data manipulation and analysis
    • PCA
  • Evaluation Techniques
    • Silhouette analysis, Davies-Bouldin index
    • 10-Fold Cross-Validation
    • Bootstrap 0.632 Method
    • Comprehensive Metrics
  • Additional: matplotlib, regex for enhanced functionality

Project Portfolio

1. 🏈 Football Data Interactive Dashboard

Objective: Interactive exploration and visualization of football dataset with real-time filtering capabilities.

Key Features:

  • Multi-perspective data analysis with dynamic filtering
  • Interactive chart generation with multiple visualization types
  • Real-time parameter adjustment for data exploration

Football Dashboard Interactive football statistics dashboard with filtering options

Football Analysis Advanced football data analysis with multiple chart perspectives

2. 🍷 Wine Quality Analysis: Regression & Clustering

Objective: Comprehensive wine quality analysis combining linear regression and K-Means clustering with interactive visualizations.

Machine Learning Techniques:

  • Linear Regression: Interactive R² calculation for all feature combinations
  • K-Means Clustering: Dynamic clustering analysis (2-5 clusters)
  • Principal Component Analysis: Dimensionality reduction for visualization
  • Evaluation Metrics: Silhouette analysis, Davies-Bouldin index

Interactive Features:

  • Real-time regression analysis with adjustable X/Y variables
  • Dynamic clustering visualization with decision boundaries
  • PCA scree plots showing explained variance ratios
  • Comprehensive statistical metrics display (R², MSE, RMSE, MAE)

Wine Analysis Interactive regression and clustering analysis with real-time parameter adjustment

3. 🚢 Titanic Survival Prediction: Cross-validation & Bootstrapping

Objective: Advanced model evaluation using cross-validation and bootstrap methods on the famous Titanic dataset.

Machine Learning Models:

  • Logistic Regression: Linear classification with probability estimates
  • Decision Tree: Rule-based classification with visual tree structures
  • K-Nearest Neighbors (k=3): Instance-based learning algorithm

Advanced Evaluation Techniques:

  • 10-Fold Cross-Validation: Robust performance estimation
  • Bootstrap 0.632 Method: Alternative evaluation with 100 iterations
  • Comprehensive Metrics: Accuracy, Precision, Recall, F1-score analysis
  • Visual Comparisons: Side-by-side confusion matrices and metric charts

Data Engineering:

  • Feature engineering with age imputation and categorical encoding
  • One-hot encoding for categorical variables
  • Systematic missing value handling

4. ⭐ Pulsar Star Classification: SVM & Neural Networks

Objective: Advanced classification of pulsar stars using Support Vector Machines and Artificial Neural Networks with comprehensive hyperparameter optimization.

Support Vector Machine Implementation:

  • Four Kernel Types: Linear, Polynomial, RBF, Sigmoid kernels
  • Interactive Parameter Tuning: C parameter (0.01-100), Gamma, Degree parameters
  • Real-time Visualizations: Decision boundary plots with PCA projection

Neural Network Architecture:

  • Model 1: Simple perceptron (8 inputs → 1 output, logistic activation)
  • Model 2: Multi-layer network (8 → 8 → 1, ReLU activation)
  • Advanced Features: Interactive topology diagrams, learning curves, weight visualization

Performance Achievements:

  • Best SVM Results: 98.08% accuracy (Polynomial & RBF kernels)
  • Neural Network Results: 98.08% accuracy (simple perceptron)
  • Technical Implementation: Grid search optimization, model persistence

Pulsar Classification SVM decision boundaries and neural network topology visualization with performance metrics

Key Technical Achievements

  • ✅ Real-time Interactivity: All applications feature dynamic parameter adjustment
  • ✅ Multiple ML Paradigms: Classification, regression, clustering, and evaluation techniques
  • ✅ High Performance: Achieved 98%+ accuracy on challenging datasets
  • ✅ Comprehensive Evaluation: Cross-validation, bootstrap methods, multiple metrics
  • ✅ Professional Visualization: Publication-quality interactive dashboards

Academic Context

Developed as the practical component for Computer Science Bachelor studies at University Leipzig, Germany. This coursework demonstrates advanced competencies in:

  • Interactive Data Visualization using modern web frameworks
  • Machine Learning Implementation with industry-standard libraries
  • Statistical Analysis with proper evaluation methodologies
  • Software Engineering with modular, documented code architecture