Objective: Predict Falcon 9 landing success
Tools: Python, SQL, Tableau
Result: Identified key predictors of success (e.g. Launch Site, Orbit) and created an interactive dashboard
This approach applies to:
- Delivery/Logistics: Predicting on-time success
- Product Reliability: Analysing production outcome logs
- Subscription Models: Customer churn pattern recognition
This project analyses SpaceX Falcon 9 launch data to uncover patterns driving mission success. It uses Python for data cleaning, exploration, and visualisation, followed by a Tableau dashboard for stakeholder insight delivery.
Techniques applied here are relevant to business use cases such as:
- Predicting delivery or production success rates
- Evaluating performance across categories (regions, vendors, teams)
- Analysing binary outcomes (success/failure, churn/stay, buy/not-buy)
See Full Report | View Dashboard
This project provides a structured analysis of SpaceX Falcon 9 launch data, focusing on historical trends, landing success probabilities, and cost estimation in commercial spaceflight. It documents the methodology, data processing, and machine learning approach used to extract key insights.
For visual trends and graphical analysis, refer to the SpaceX Data Visualisation Report.
- Python: Employed for data collection, cleaning, and predictive modelling using libraries such as Pandas, Matplotlib, Seaborn, and Scikit-Learn.
- SQL: Used for structured data querying and statistical aggregation.
- Tableau: An interactive dashboard was developed to visualise trends and insights.
The commercial space industry has undergone rapid transformation, with companies such as SpaceX, Blue Origin, and Rocket Lab driving innovation. SpaceX, in particular, has pioneered the reuse of Falcon 9 first-stage boosters, reducing launch costs significantly. This analysis examines historical launch data to determine factors influencing successful booster landings, providing insights relevant to cost estimation and mission planning.
This project seeks to determine the probability of Falcon 9 first-stage recovery through a structured analytical approach:
- Data Wrangling & Cleaning: Missing values were handled, and dataset integrity was ensured.
- Exploratory Data Analysis (EDA): Statistical analysis and visualisation techniques were employed to extract key insights.
- Machine Learning Model: A predictive model was trained to classify landing success.
- Interactive Dashboard: A Tableau dashboard was created to facilitate exploratory analysis.
Two primary sources were utilised:
- SpaceX API: Provides structured data on Falcon 9 missions, booster reusability, and mission outcomes.
- Wikipedia (Snapshot: 9 June 2021): A static dataset compiled from historical launch records.
The datasets were processed to ensure uniform formatting, removal of inconsistencies, and the integration of relevant attributes for analysis.
A statistical and graphical investigation was conducted, focusing on:
- Launch success rates across different time periods.
- Booster reuse trends and their correlation with mission outcomes.
- Payload mass distributions in relation to landing success.
- Geospatial patterns associated with launch sites and landing locations.
SQL queries were executed to extract structured insights from the dataset:
- Launch success rates by site and booster version.
- Payload mass statistics across different mission types.
- Booster performance trends, including recovery success probabilities.
- Aggregated insights, such as total payload delivered for NASA CRS missions.
A classification model was trained to predict booster recovery success. The modelling process involved:
- Supervised Learning Techniques: Logistic regression, decision trees, support vector machines (SVM), k-nearest neighbours (KNN), and random forests.
- Feature Engineering: Selection of key predictors, including payload mass, launch site, and booster version.
- Performance Evaluation: Metrics such as accuracy, precision, recall, and F1-score were used to assess model effectiveness.
Static and interactive visualisations were used to enhance analytical insights:
- Descriptive Statistics: Histograms, box plots, and correlation matrices.
- Geospatial Analysis: Launch site distributions and landing outcome visualisations.
- Interactive Tableau Dashboard: Available here.
For an in-depth analysis of key visual trends, refer to the SpaceX Data Visualisation Report.
- Falcon 9 carried an average payload of 6,130.55 kg, demonstrating its operational versatility.
- NASA CRS missions transported 47,094.7 kg of cargo, highlighting the rocket’s role in ISS resupply.
- Launch successes improved significantly from 2014 onward, reaching an 80% success rate by 2020.
- Logistic regression provided the highest in-sample accuracy (94.4%), whereas decision trees demonstrated better generalisation.
- Payload mass alone was not a strong determinant of landing success; booster version and flight history were more significant predictors.
- Kennedy Space Center LC-39A and Cape Canaveral SLC-40 recorded the highest success rates.
- Payloads above 10 tons exhibited a lower probability of successful recovery.
- Controlled landings became more consistent post-2016, coinciding with improvements in reusability technology.
This analysis provides insights into the factors influencing Falcon 9 booster recovery success. The findings are relevant for cost estimation, mission planning, and reusable rocket optimisation in commercial spaceflight.
- Python 3.12
- Jupyter Lab
- PostgreSQL
- Tableau Public
- Required dependencies can be installed using:
pip install -r requirements.txt
or, for Conda users:
conda env create -f environment.yml
For further inquiries or collaboration opportunities, please get in touch via the following channels:
- Author: Josué Gómez Parada
- Email: josue[dot]gomez[dot]parada[at]gmail[dot]com
- LinkedIn: dr-JGP
- GitHub: jgp-13
- Tableau Public: Visualisations & Reports
- Website: Portfolio & Background
This project and its contents are licensed under the MIT Licence, unless otherwise stated.
Copyright © 2021 IBM Corporation. All rights reserved.
Unauthorized use, reproduction, or distribution of this material without explicit permission is prohibited.