🚀 Startup Success Prediction

This repository contains a data mining project developed as part of the IN015 - Data Mining course.
The goal is to predict startup success using real-world data and machine learning techniques.

---

📂 Project Overview

Startups are inherently risky, but certain indicators—such as social media activity, engagement ratios, and web presence—may help assess their probability of success.
In this project, we:

• Performed data cleaning and preprocessing on a dataset of 700 startups.
• Handled missing values with domain-based imputations.
• Applied exploratory data analysis (EDA), correlation heatmaps, and visualizations.
• Built and compared multiple machine learning models:
  • Dummy Classifier (baseline)
  • Logistic Regression
  • Decision Tree
  • Random Forest
  • XGBoost

Our best-performing model was Random Forest, achieving ~88.9% precision in predicting successful startups.

---

🛠️ Tech Stack

• Languages: Python 3
• Libraries:
  • Data Handling: pandas, numpy
  • Visualization: matplotlib, seaborn
  • Machine Learning: scikit-learn, xgboost
  • Evaluation: metrics, cross-validation

---

📊 Results

• Dummy Classifier: Very poor (predicts all startups fail).
• Logistic Regression: ~75% precision.
• Decision Tree: Weak performance (overfits easily).
• Random Forest: Best model with 88.9% precision.
• XGBoost: Promising results, comparable to Random Forest.