IBM Data Science Capstone Project Space X

Jennyfer WAN

Winning Space Race with Data science

Executive Summary

Jennyfer WAN

Appendix

Conclusion

Methodology

Results

Introduction

Outline

Jennyfer WAN

executive summary

Jennyfer WAN

1. Exploratory Data Analysis results
2. Interactive analytics in screenshots
3. Predictive Analytics results

Summary of all results

1. Data Collection through API
2. Data Collection with Web Scraping
3. Data Wrangling
4. Exploratory Data Analysis with SQL
5. Exploratory Data Analysis with Data Visualization
6. Interactive Visual Analytics with Folium
7. Machine Learning Prediction

Summary of methodologies

Executive Summary

Jennyfer WAN

introduction

Jennyfer WAN

1. What factors determine if rockets will land successfully?
2. Which features are the most correlatedto determine the success rate of a successful landing.
3. What conditions does SpaceX have to achieve to get the best results and ensure the best rocket success landing rate.

Problems you want to find answers

SpaceX advertises Falcon 9 rocket launches on its website with a cost of 62 million dollars; other providers cost upward of 165 million dollars each, much of the savings is because SpaceX can reuse the first stage. Therefore if we can determine if the first stage will land, we can determine the cost of a launch. This information can be used if an alternate company wants to bid against SpaceX for a rocket launch. In this capstone, we will predict if the Falcon 9 first stage will land successfully using the machine learning pipeline created.

Project background and context

Introduction

Jennyfer WAN

Methodology

Jennyfer WAN

Data collection methodology :

• SpaceX Rest API.
• Web Scraping from Wikipedia.

Perform data wrangling :

• One Hot Encoding applied on categorical features (Transforming data for Machine Learning)

Perform exploratory data analysis (EDA) using visualization and SQL :

• Scatter and bar graphs to show patterns between data.

Perform interactive visual analytics using Folium and Plotly Dash Perform predictive analysis using classification models :

• How to build, tune and evaluate classification models.

Methodology

Jennyfer WAN

Create dataframe from dictionnary

Apply list to dictionnary

Filter columns then export csv

Clean data, check and fill missing values

Convert into dataframe using .json_normalize()

Decode the response content as a Json using .json()

Request to get SpaceX API

Data collection is the process of gathering data to provide the information that's needed to answer questions, analyze business performance or other outcomes, and predict future trends or actions to take.

Data collection - SpaceX api

Jennyfer WAN

Create a dictionnary

Extract column name one by one

Convert dictionnary to dataframe then export csv

Appending data to keys

Find all tables

Create a BeautifulSoup object

Request the Falcon9 Launch HTML page

Data collection is the process of gathering data to provide the information that's needed to answer questions, analyze business performance or other outcomes, and predict future trends or actions to take.

Data collection - web scraping

Jennyfer WAN

Export dataframe to csv

Create a landing outcome label from Outcome column

Calculate the number and occurence of mission outcome per orbit type

Calculate the number and occurrence of each orbit

We perform Exploratory Data Analysis (EDA) and determine Training Labels We convert those outcomes into Training Labels with : - 1 means the booster successfully landed - 0 means it was unsuccessful

Calculate the number of launches on each site

Data wrangling is the process of cleaning and unifying messy and complex data sets for easy access and analysis.

Data wrangling

Jennyfer WAN

Line graphs work well in showing trends chronologically.Moreover, we can visualize data changes at a glance.

Most audiences understand how to read a bar graph and can grasp the information.From this graph, we can easily interpret which orbit have the highest sucess rate

We use a scatter plot to determine whether or not two variables have a relationship or correlation.

• Launch Success Yearly Trend

• Success rate and Orbit type

• FlightNumber and PayloadMass
• Flight Number and Launch Site
• Payload and Launch Site
• FlightNumber and Orbit type
• Payload and Orbit type

Line Graph

Bar Graph

Scatter Graphs

Exploratory data analysis is an approach to analyzing data sets to summarize their main characteristics by using simple tools from statistics, simple plotting tools.

EDA with Data Visualization

Jennyfer WAN

EDA with SQL (Structured Query Language)

We were able to load SpaceX dataset into the corresponding table in a Db2 database directly on Jupyter notebook. We performed EDA with SQL queries to gather information from the data :

• Display the names of the unique launch sites in the space mission
• Display 5 records where launch sites begin with the string 'CCA'
• Display the total payload mass carried by boosters launched by NASA (CRS)
• Display average payload mass carried by booster version F9 v1.1
• List the date when the first successful landing outcome in ground pad was acheived.
• List the names of the boosters which have success in drone ship and have payload mass greater than 4000 but less than 6000
• List the total number of successful and failure mission outcomes
• List the names of the booster_versions which have carried the maximum payload mass. Use a subquery
• List the failed landing_outcomes in drone ship, their booster versions, and launch site names for in year 2015
• Rank the count of landing outcomes (such as Failure (drone ship) or Success (ground pad)) between the date 2010-06-04 and 2017-03-20, in descending order

It is the standard language to interact with databases. SQL is the most important tool, a data analyst uses to manipulate and gain insights from the data.

Jennyfer WAN

Folium makes it easy to visualize data that has been manipulated in Python on an interactive leaflet map which makes it an excellent tool for plotting maps.

Build an Interactive Map with Folium

In order to make an interactive visual analytics :

• We marked all launch sites on a map and added map objects such as :
• folium.Marker()
• folium.Circle()

• We assigned the feature "launch_outcome" to easily visualize marker colors on the map based on the class value :
• 1 (Success) = Green
• 0 (Failure) = Red

• We marked the success/failed launches for each site on the map with folium.Icon() :
• The color-labeled marker clusters, able us to easily identify which launch sites have relatively high success rates.

• We calculated the distances between a launch site to its proximities thanks to MousePosition to get coordinate for a mouse over a point on the map.

• Then we answered some questions :
• Are launch sites in close proximity to railways, highways and coastline ?
• Do launch sites keep certain distance away from cities ?

Jennyfer WAN

- Add a Scatter Plot to show the relationship with Outcome and Payload Mass for the different Booster Version.- Scatter plot is used to determine whether or not two variables have a relationship or correlation.

Scatter Plot :

03

- Add a Range Slider to Select Payload Mass (Kg) in order to find if variable is correlated to mission outcome.

Range Slider :

02

01

- Add a Launch Site Dropdown Input Component to show the total launches by all sites or a certain site.- Generally used to display numeric values, Pie chart is easy to understand thanks to its different portions and color codings.

Pie Chart :

Dash is a python framework created by plotly for creating interactive web applications.

Build a Dashboard with Plotly Dash

Jennyfer WAN

• Select the model with the best accuracy score

Predictive analytics is the use of various statistical and machine learning algorithms to predict the likelihood of future outcomes based on historical data. The goal is to suggest a course of action or strategy to make decisions from immediate to long term to provide the best assessment of what will happen in the future.

Predictive Analysis (Classification)

• Use the accuracy metric for our model

• Improve our model by using feature engineering and algorithm tuning to find the best hyperparameters for each type of algorithms

• Plot a Confusion Matrix

2. Evaluating Model

3. Find the best performing Classification Model

• Load our dataset using NumPy and Pandas
• Transform our data
• Split our data into train and test set
• Check how many test samples have been created
• Select the different machine algorithms to be trained
• Set hyperparameters and algorithms to the object GridSearchCV

• Fit the the data into the GridSearchCV to find the best parameters

• Train our dataset

1. Building Models

Jennyfer WAN

results

Jennyfer WAN

Results

• Exploratory data analysis results

• Interactive analytics demo in screenshots

• Predictive analysis results

Jennyfer WAN

• Flight Number vs. Launch Site
• Payload vs. Launch Site
• Success Rate vs. Orbit Type
• Flight Number vs. Orbit Type
• Payload vs. Orbit Type
• Launch Success Yearly Trend

EDA WITH visualization

Jennyfer WAN

Flight Number vs. Launch Site

As we can see from this scatter plot, the more the number of flights increases, the more the success rate increases

Jennyfer WAN

Payload vs. Launch Site

For example with CCAFS SLC 40 LaunchSite, the greater the playload mass, the greater the success rate for the rocket to land.At this stage, we still cannot determine if there is a correlation between these two variables.

Jennyfer WAN

Success Rate vs. Orbit Type

From the barplot, we can see that 'ES-L1', 'GEO', 'HEO', 'SSO' had the most success rate.

Jennyfer WAN

Flight Number vs. Orbit Type

We can clearly observe for 'LEO' Orbit that sucess is related to the number of flights.Unlike 'GTO' where there seems to be no correlation.

Jennyfer WAN

Payload vs. Orbit Type

With heavy payloads, the successful landing are more for 'PO', 'ISS' and 'LEO' orbits.Unlike 'GTO' which have a negative impact.

Jennyfer WAN

Launch Success Yearly Trend

From this line chart, we can observe that the success rate keep increasing since 2013 till 2020.

Jennyfer WAN

• All Launch Site Names
• Launch Site Names Begin with 'CCA'
• Total Payload Mass
• Average Payload Mass by F9 v1.1
• First Successful Ground Landing Date
• Successful Drone Ship Landing with Payload between 4000 and 6000
• Total Number of Successful and Failure Mission Outcomes
• Boosters Carried Maximum Payload
• 2015 Launch Records
• Rank Landing Outcomes Between 2010-06-04 and 2017-03-20

EDA WITH SQL

Jennyfer WAN

RESULT

SQL QUERY

We used the keyword DISTINCT to show unique values in LaunchSite column from SpaceXTbl dataset

All Launch Site Names

Jennyfer WAN

➜ Display first 5 records where column Launch_Site values must start with 'CCA'

SELECT * FROM = Query all (*) rows and columns from SpaceXTbl dataset WHERE ... LIKE ... = Condition which will only query rows from Launch_Siteusing pattern matching % LIMIT = Return first n_rows matching the SELECT criteria.

RESULT

SQL QUERY

Launch Site Names Begin with 'CCA'

Jennyfer WAN

➜ Display the total payload mass carried by boosters launched by NASA (CRS)

SELECT SUM( ) = Return the total of the column payload_mass__kg_ AS = rename the name of column WHERE ... = ... = Condition which will only query rows from Customer columns with 'NASA (CRS)' values

RESULT

SQL QUERY

Total Payload Mass

Jennyfer WAN

➜ Display average payload mass carried by booster version F9 v1.1

SELECT AVG( ) = Return the average of the column payload_mass__kg_ WHERE ... = ... = Condition which will only query rows from Booster_version column with 'F9 v1.1' values

RESULT

SQL QUERY

Average Payload Mass by F9 v1.1

Jennyfer WAN

➜ Display the date when the first successful landing outcome in ground pad was acheived.

SELECT MIN( ) = Return the average of the column payload_mass__kg_ WHERE ... = ... = Condition which will only query rows from Launding__outcome column with 'Success (ground pad)' values

RESULT

SQL QUERY

First Successful Ground Landing Date

Jennyfer WAN

➜ Display the names of the boosters which have success in drone ship and have payload mass greater than 4000 but less than 6000

SELECT = Query only data from booster_version column WHERE ... = ... = Condition which will only query rows from Launding__outcome column with 'Success (drone ship)' values AND = Requires that additional conditions are true

RESULT

SQL QUERY

Successful Drone Ship Landing with Payload between 4000 and 6000

Jennyfer WAN

➜ Display the total number of successful and failure mission outcomes

SUM(CASE WHEN ... THEN ... ELSE ... END) = Case statement is used to get both success and failure instead of multiple COUNT()

RESULT

SQL QUERY

Total Number of Successful and Failure Mission Outcomes

Jennyfer WAN

➜ Display the names of the booster_versions which have carried the maximum payload mass.

Here, we used a subquery for :WHERE ... = (SELECT MAX( ) FROM ...) = return the maximum value of payload_mass__kg_ column

RESULT

SQL QUERY

Boosters Carried Maximum Payload

Jennyfer WAN

➜ Display the failed landing_outcomes in drone ship, their booster versions, and launch site names for in year 2015

MONTHNAME(...) = return the month from the column Date WHERE ...AND ... = requires all conditions to be true. Here, we only select Failure (drone ship) from 2015

RESULT

SQL QUERY

2015 Launch Records

Jennyfer WAN

• All Launch Sites on folium map
• Color labels for each site on the map
• Launch Sites distances from railways / highways / cities / coastlines

LAUNCHSITES PROXIMITIES analysis

Jennyfer WAN

California

Florida

➜ Launch sites are close proximity to the coast for safety reasons

All Launch Sites on folium map

Jennyfer WAN

Green Markers = Successful Launches Red Markers = Failure Launches ➜ KSC LC-39A launch site has the most probability of success

Color labels for each site on the map

➜ Do launch sites keep certain distance away from cities? Launch sites are the farthest from cities and dense areas to protect the population from them.

➜ Are launch sites in close proximity to coastline ? Launch sites are close to coastline for multiple logical reasons : - As we saw on previous notebooks, the launch success rate may depend on many factors such as the location and proximities of a launch site, i.e., the initial position of rocket trajectories. - Do the lauches over the ocean to cancel any time in case of problems. - Prevent human and material repercussions in case of failure.

➜ Are launch sites in close proximity to highways ? Launch sites still close to highways for the same reasons for railways. But since the highways are also frequented by the population, they must keep a safe distance to avoid any injuries.

Closest_Coastline

Closest_City

Closest_Highlway

Closest_Railway

➜ Are launch sites in close proximity to railways ? Launch sites are nearest from railways in order to transport and receive more easily materials or cargos. But also, to minimize the distance for the employees : thus saving time, money and effort.

Launch Sites distances from railways / highways / cities / coastlines

Jennyfer WAN

• Success Count for all Launch Sites with pie chart
• Pie chart with highest success ratio
• Folium Map Screenshot 3

build a dashoboard with plotly dash

Jennyfer WAN

Has we saw on Folium Map part, KSL LC-391 had the most successeful launches from Launch Sites.

Success Count for all Launch Sites with pie chart

Jennyfer WAN

KSL LC-39A had 76.9% of success rate while getting 23.1% of failure rate

pie chart with highest success ratio

(with different Payload selected in Range slider) PART I

Payload range(Kg) between 0 to 10 000 Kg

Payload range(Kg) between 0 to 5600 Kg

Jennyfer WAN

Low weighted Payload Mass (Kg) have HIGHER success rate than Heavy Payload Mass (Kg)

scatter plot of Payload vs launch outcome for all sites

Booster Version Company with Highest success rate

Payload range(Kg) with highest success rate

(with different Payload selected in Range slider) PART II

Payload range(Kg) with lowest success rate

Jennyfer WAN

Payload range with lowest success rate is between 362 Kg and 475kg. Most payload mass with highest success rate is between 1952 Kg and 5300kg. FT is the Booster Version with highest launch success rate.

scatter plot of Payload vs launch outcome for all sites

Jennyfer WAN

• Classification Accuracy
• Confusion Matrix

predictive analysis (classification)

Jennyfer WAN

We trained 4 models different models. Decision Tree has the highest classification accuracy with 0.90 (while during test set, it got 0.83, the lowest score)

Classification Accuracy

Same Confusion Matrix for KNN, Decision Tree and Logistic Regression

False Negative

+ INFO

Jennyfer WAN

Decision Tree got a higher result thanks to his TP of 3 against TP of 5 for the other models.On the other hand, it calculates more than FN (3 against 1).

Confusion matrix

Jennyfer WAN

conclusion

Jennyfer WAN

• More the number of flights increases, more the success rate increases at a launch site.

• Orbits 'ES-L1', 'GEO', 'HEO', 'SSO' had the most success rate.

• Success rate keep increasing since 2013 till 2020.

• KSC LC-39A had the most successful launches from all sites.

• Low payloads mass (Kg) perform better than the heavier payloads.

• Most payload mass with highest success rate is between 1952 Kg and 5300kg.

• FT is the Booster Version with highest launch success rate.

• The Decision Tree classifier is the best machine learning algorithm for this project with provided dataset.

Conclusions

Jennyfer WAN

appendix

Jennyfer WAN

Interactive Plotly : https://plotly.com/python-api-reference/ Dash Plotly :https://dash.plotly.com DashBoarding Tools : https://pyviz.org/dashboarding/

LINKS

Jennyfer WAN

THANKS!