🚀 SpaceX Falcon 9 Landing Success Prediction

IBM Data Science Professional Certificate – Capstone Project

📌 Project Overview

The goal of this project is to predict whether the Falcon 9 first stage will successfully land, a critical factor in reducing launch costs and maintaining SpaceX’s competitive advantage in the space industry.

Because SpaceX can reuse boosters after a successful landing, the ability to predict landing outcomes is valuable for financial planning, mission design, and risk reduction.

This end-to-end data science project includes:

• Data collection via API + web scraping
• Feature engineering & data cleaning
• Exploratory Data Analysis (EDA)
• Interactive visual analytics
• Machine learning model development
• Model evaluation & insights

---

🎯 Business Problem

SpaceX markets Falcon 9 launches at ~$62M, but the actual marginal cost can drop to under $30M due to booster reuse.

Being able to predict landing success helps:

• Reduce operational risk
• Improve launch planning
• Lower mission cost uncertainty
• Support pricing strategy for customers

---

📂 Dataset

The dataset includes historic Falcon 9 launches and contains features such as:

• Flight Number
• Launch Site
• Orbit Type
• Payload Mass
• Booster Version
• Flight Outcome
• Launch Site Coordinates
• Landing Outcome (Target Variable: Success/Failure)

Data sources:

• SpaceX REST API
• Web scraping (SpaceX Wikipedia page)
• Provided CSV datasets from IBM Skills Network

---

🔍 Exploratory Data Analysis (EDA)

Key analytical steps:

⭐ Univariate & Bivariate Analysis

• Examined landing success rates across launch sites
• Analyzed payload mass vs. landing outcome
• Explored orbit type impact on landing probability
• Investigated booster version patterns

⭐ Geographic Visualization

• Mapped launch sites & landing locations using Folium

⭐ Interactive Dashboards

(If included in your version)

• Created Dash app with launch site filters
• Displayed payload vs. class scatter plot
• Showed success rate by launch site

---

🤖 Machine Learning Approach

Models Trained

• Logistic Regression
• Support Vector Machine (SVM)
• K-Nearest Neighbors (KNN)
• Decision Tree Classifier

Feature Engineering

• Created dummy variables
• Normalized continuous features
• Tuned hyperparameters using GridSearchCV

---

🧪 Model Evaluation

Metrics used:

• Accuracy Score
• F1 Score
• Confusion Matrix

Best Model:

🎉 The Decision Tree Classifier (or whichever performed best in your results) achieved the highest accuracy and balanced performance between precision and recall.

This model effectively captures nonlinear relationships present in the SpaceX launch data.

---

📈 Key Insights

• Launch Site KSC LC-39A has one of the highest landing success rates.
• Heavier payloads slightly reduce the probability of a successful landing.
• Newer booster versions (e.g., Block 5) drastically improve landing success.
• Orbits such as GTO are more challenging and reduce success probability.
• Machine learning confirms: launch site + booster version are the strongest predictors.

---

💼 Business Recommendations

1️⃣ Optimize Launch Scheduling

Prioritize launches from sites with proven high success rates to reduce risk.

2️⃣ Invest in Booster Technology

Block 5 and newer boosters significantly outperform earlier versions — continued investment decreases failure rate and cost per launch.

3️⃣ Use ML Predictions in Mission Planning

Integrate this model into SpaceX’s planning workflow to forecast landing success before each mission.

4️⃣ Customer Pricing Strategy

Higher predicted landing probability → lower operational risk → opportunities for more competitive pricing.

---