https://github.com/bibymaths/ifa-2022

Computational Algorithms & Data Science Projects
https://github.com/bibymaths/ifa-2022

bioinformatics machine-learning modelling pattern-matching

Last synced: about 1 year ago
JSON representation

Computational Algorithms & Data Science Projects

• Host: GitHub
• URL: https://github.com/bibymaths/ifa-2022
• Owner: bibymaths
• License: mit
• Created: 2023-10-10T07:35:22.000Z (over 2 years ago)
• Default Branch: main
• Last Pushed: 2025-01-31T22:24:20.000Z (over 1 year ago)
• Last Synced: 2025-02-11T12:00:01.340Z (over 1 year ago)
• Topics: bioinformatics, machine-learning, modelling, pattern-matching
• Language: C++
• Homepage:
• Size: 60.4 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# **Group 7 - Computational Algorithms & Data Science Projects**

## **Overview**

This repository contains various projects developed as part of the **Introduction to Focus Areas WS 22/23** at Freie Universität Berlin. The projects explore advanced algorithms, data science, machine learning, and computational techniques applied in real-world biological and medical datasets.

## **Projects & Reports**

### **1. Advanced Algorithms - Search Implementation**

- **Report:** [`here`](Reports/Advanced_Algorithms.pdf)

- **Poster:** [`here`](Poster/Poster_Advanced_Algorithms.pdf)

- **Programming Languages:** C++, Perl, R

- **Description:** Implemented and benchmarked various search algorithms:

  - **Naive Search**

  - **Suffix Array Search**

  - **FM-Index Search**

  - **FM-Index Search with Pigeonhole Principle**

  - **Performance evaluation (runtime & memory usage)**

- **Key Findings:** FM-Index search was the most efficient for large datasets, while naive search was computationally expensive.

### **2. Complex Systems - Viral Infection Modeling**

- **Report:** [`here`](Reports/Complex_Systems.pdf)

- **Programming Language:** Python

- **Description:**

  - Developed and simulated **ordinary differential equation (ODE) models** to study viral infections.

  - Performed **stochastic simulations** with the Stochastic Simulation Algorithm (SSA).

  - Estimated unknown parameters using **optimization methods**.

  - Analyzed how **infection probability changes** with viral exposure.

- **Key Findings:** Identified key parameters affecting infection dynamics and established relationships between viral exposure and infection probability.

### **3. Data Science - Classification & Machine Learning**

- **Report:** [`here`](Reports/Data_Science.pdf)

- **Presentation:** [`here`](Presentation/presentation.pdf)

- **Programming Language:** R, Python

- **Description:**

  - **Project 1:** Classification of heart disease using machine learning models.

    - Logistic Regression, Random Forest, Boosted Logistic Regression, k-Nearest Neighbors

    - Performance metrics: Accuracy, Sensitivity, Specificity, AUC-ROC

  - **Project 2:** Breast cancer classification using deep learning models.

    - Convolutional Neural Network (CNN)

    - Fully Connected Neural Network

    - Shallow Neural Network (SNN)

- **Key Findings:** CNN outperformed other models for breast cancer classification, while Boosted Logistic Regression had the best accuracy for heart disease classification.

#### **Breast Cancer Image Classification**

- **Programming Language:** Python (TensorFlow, Scikit-learn)

- **Dataset:** BreaKHis - 7909 microscopic images from breast cancer tissues

- **Description:**

  - Implemented deep learning models to classify **benign vs. malignant** tissues.

  - Used **CNN, Fully Connected NN, and Shallow NN** for classification.

  - Evaluated models with **confusion matrices, accuracy plots, ROC curves**.

- **Key Findings:** CNN performed best, while SNN and Fully Connected NN suffered from imbalanced training data.

## **Installation & Dependencies**

To run the code from the Jupyter notebooks or scripts, ensure you have the following installed:

```bash

pip install numpy pandas matplotlib seaborn tensorflow scikit-learn

```

For R-based projects, install the necessary libraries:

```r

install.packages(c("caret", "ggplot2", "randomForest", "MLeval"))

```

## **Contributors**

- **Abhinav Mishra** (Bioinformatics) – Algorithm development, data science, and optimization.

- **Jule Brenningmeyer** (Bioinformatics) – Machine learning and stochastic modeling.

- **Maike Herkenrath** (Bioinformatics) – Data preprocessing and deep learning.

- **Se Yeon Kim** (Data Science) – Heart disease classification and neural network training.

## **References**

- Detrano et al. (1989). *International application of a new probability algorithm for coronary artery disease*.

- Spanhol et al. (2016). *A Dataset for Breast Cancer Histopathological Image Classification*.

- Reinert, K. *FM-Index and Suffix Array Algorithms*