Inspiration
We do not know enough about the future. It is scary and unknown, yet we try to fight the irregularities that happen in our everyday lives.
Thus we must predict it.
Predicting the failing rates would help a lot in decreasing measurement time.
What it does
It predicts the breakdown probability and the best measurements to do in order to reject the device as early as possible. This model learns from the past device measurements and uses the currently known characteristic to predict the value of each measurement to be done and order them, failing probability ascending order.
How I built it
We wrote lines of python code, and used our great machine learning knowledge to create model of the device's measurements.
During this project we used different libraries from sklearn and pandas model using different machine learning algorithms to create proof of concepts.
Challenges I ran into
Time and Money, for some reason we ran out of both too soon.
We had some challenges when trying to run our algorithms on our machines as we would have preferred more extensive tests. We also believe that in real data correlation could be a lot more visible and would improve the use of PCA in our model.
Accomplishments that I'm proud of
We are proud to announce that the future is now as known as the past.
What I learned
We learned that breakdowns are rare but present.
What's next for breakdown_prediction
There will be no more breakdowns so we will be jobless.
We hope to test our algorithms on real data and improve our implementation to not only be used as a POC but also as an actual predictive algorithm.
Log in or sign up for Devpost to join the conversation.