IBM Data Science Course

I embarked on a journey to prepare for a career in data science, acquiring in-demand skills and hands-on experience to become job-ready.

IBM Data Science Professional Certificate

What I Learned

Throughout this comprehensive program, I gained proficiency in:

• Data Science: Acquiring the fundamental concepts and methodologies of data science.
• Methodology: Understanding the systematic approach to data analysis.
• Machine Learning: Developing expertise in machine learning algorithms.
• Python Programming: Becoming proficient in Python, a key language in data science.
• Big Data: Handling and analyzing large datasets.
• Data Mining: Extracting valuable insights from data.
• SQL: Mastering SQL for database management.

Notable Projects

During the program, I completed several projects that showcased my data science proficiency:

• Extracting and visualizing financial data using the Pandas Python library.

• Querying census, crime, and school demographic data sets using SQL.

• Wrangling data, creating graphs, and building regression models to predict housing prices with Python data science libraries.

• Developing a dynamic Python dashboard for monitoring and improving US domestic flight reliability.

• Applying machine learning classification algorithms to predict loan repayment.

• Training and comparing machine learning models to predict the reuse of the first stage of a rocket in space launches.

Capstone Project

SpaceX Launch Success Analysis

Overview

This study investigates the determinants of launch success in SpaceX's rocket endeavors. As an industry leader, SpaceX endeavors to democratize space travel. Through innovative reuse of the Falcon 9 rocket's first stage, they have significantly reduced launch costs to approximately $62 million per launch, providing a stark contrast to competitors whose costs often exceed $165 million per launch. The accurate prediction of first-stage landings is of paramount importance in determining launch expenses. This analysis harnesses publicly available data and machine learning techniques to predict the reusability of the first stage, both for SpaceX and its competitors.

Exploration

Our investigation delves into the influence of several factors on first-stage landing success, including payload mass, launch site, number of flights, orbits, and temporal success rate trends. The primary objective is to identify the most effective predictive model via binary classification.

Executive Summary

This research follows a rigorous methodology:

Data Collection

Data collection is carried out through SpaceX's REST API and web scraping procedures. A critical component of data wrangling involves the creation of a success/failure outcome variable.

Data Exploration

Data visualization scrutinizes payload characteristics, launch site dynamics, flight numbers, and temporal patterns. Moreover, SQL queries yield pivotal statistics, such as cumulative payload mass, payload mass ranges for successful launches, and the total count of successful and failed outcomes.

Data Analysis

We meticulously assess launch site success rates and their proximity to geographic markers, concurrently visualizing sites with notable success rates and payload ranges.

Predictive Modeling

Leveraging logistic regression, support vector machine (SVM), decision tree, and K-nearest neighbor (KNN) models, we predict landing outcomes and discern the model with the highest predictive capability.

Key Findings

Exploratory Data Analysis

• Success rates have demonstrated an upward trajectory over time.
• Kennedy Space Center's Launch Complex 39A (KSC LC-39A) consistently attains the highest success rate among landing sites.
• Specific orbits, namely ES-L1, GEO, HEO, and SSO, consistently achieve a 100% success rate.

Data Visualization and Analytics

A strategic placement of launch sites in proximity to the equator optimizes Earth's rotational speed. All launch sites are situated in coastal regions.

Predictive Analytics

Model performances on the test set are quite homogeneous, with the decision tree model slightly outperforming its counterparts.

Methodology

Our comprehensive approach encompasses:

• Data collection via SpaceX REST API and web scraping methodologies.
• Data wrangling, including the creation of a binary success/failure variable.
• Data exploration through visualization and SQL queries.
• Cartographic visualization using Folium to portray launch sites, outcomes, and proximity measurements.
• Dashboard creation via Plotly Dash.
• Predictive analytics utilizing a suite of models and performance evaluation.

Conclusion

Key takeaways from our analysis include:

• Model performances are relatively consistent, with a slight advantage for the decision tree model.
• The strategic placement of launch sites near the equator and coastlines is prominent.
• A discernible positive trend is observed in launch success rates over time.
• Kennedy Space Center's Launch Complex 39A (KSC LC-39A) emerges as a standout performer.
• Specific orbits, including ES-L1, GEO, HEO, and SSO, consistently attain a 100% success rate.
• Payload mass exerts a notable influence on launch success, with heavier payloads correlating positively with success rates.