Dangerous Airlines & Why do planes crash “a text analysis”
Weather, terrorism, mechanical failure and human error can all bring down a plane. Which airlines crash the most?
Today I looked into about 5000 crashes recorded between 1908 and 2009 to gain insights for informed decisions on carrier choices.
This is what I found:
- 1960s-1970s are pretty bad years
- Air Taxi (small commercial aircrafts for on-demand short flights) have the most accidents
- Lufthansa, Air France, China National Aviation Corporation all used to have high accidents records but got safer over the years
Since aircrafts have different sizes, more accident doesn’t necessarily mean more fatalities.
- Aeroloft, a Russian carrier, have the highest fatality count but a relatively clean record 1990s onwards
- Several American airlines made the list: American Airlines, United and Pan Am which bankrupted and partly assimilated into Delta
However, these 10 airlines constitutes only 20% of overall fatalities.
Majority of carriers has only 1 crash. Full code on my github.
A text analysis on why do planes crash
I explore text mining to examine reasons for plane crashes.
The frequent terms revel useful keywords related to causes like pilot, fire and fog. It also contains another set of words description flying phases: take-off, landing, en-route.
Looking into words associated with most frequent keywords will give a clearer picture. As a word could be associated with multiple words, they can be viewed on a network.
Now we can make sense of the keywords with interactive graph: fire from overheating, fuel starvation, poor visibility as well as poor judgement, low altitude, overrun the runway, navigational error, and so on.
What I learnt today:
Using wordcloud2 package, one can turn term frequency into wordcloud masked by custom images or letters. However, one need to be mindful of font size, else top words may not show up.
Full code on my github.
Thanks for reading and welcome to send me ideas and suggestions.
