Predicting Supersonic Planes

Inspiration

I am excited about the Boom Supersonic Overture aircraft being developed right now. I wanted to do something machine learning related in any case to try it out, so being excited about that I wanted to incorporate supersonic aircraft into my project.

What it does

This piece of software takes a set of airplane objects that are either supersonic or not supersonic, and learns to predict whether or not an airplane is a supersonic aircraft based on a randomly generated airspeed each is given inside unique parameters.

How we built it

I built this by constructing an airplane class in which to define and manipulate airplane objects, a machine class which contains all of the helper methods, and a main class that holds the main method. I used the Bernoulli Naive Bayes model for my machine learning which I got from the sci-kit learn library.

Challenges we ran into

There were a lot of challenges to tackle in this one.

The first was figuring out what machine learning was and getting the grasp of just the most basic model, I started a crash course on machine learning and after about 7 hours got to know enough to understand the basics.

The second was figuring out how to take the data I made by creating arrays of plane objects and arrays of their attributes and put it into this model.

The second biggest one though was debugging it enough so that it would actually run something.

Then there is the one I haven't solved yet. Once the code ran, the model was only predicting everything as false. Unfortunately the function running the model is something I will be unable to manipulate for this project so the biggest challenge was building my own Bernoulli Naive Bayes algorithm based on the data I want to put into it.

Accomplishments that we're proud of

Getting code to run pretty accurately was the biggest accomplishment I am proud of. I am also very happy with the fact that I had finished the machine learning crash course on ml expert and even understood one model!

What we learned

I learned that machine learning is hard and involves a lot of mathematics. It was also a learning experience trying to implement this model with data I wanted to use. Moreover it was huge to recognize what kind of data is used for this model and how it all fits in. It was so much good experience that I am so glad that I tried this even if there is still work that needs to be done.

What's next for Predicting Supersonic Planes

What's next is optimizing the accuracy. I am a bit concerned that the threshold to get maximum accuracy is so low and remains constant as the number climbs. It may be a bug or just the nature of the range of speed for the supersonic planes. What will be further is investigating it.

Built With

• pickle
• python