Inspiration
Inspired to attend the first datathon held by MLH, we came across a variety of different projects. In particular the Conoco Philips Challenge of predicting hardware equipment failure intrigued our team. After talking through the Challenge we decided the problem would be interesting to solve given our current skills.
What it does
Our solution is a software that takes data from the sensors of hardware failures, and uses that data and feeds it into our trained classifier model. From there we can predict where the equipment failure could likely occur and furthermore reduce the costs of such failures.
How we built it
We built the program using python and implemented a series of machine learning libraries such as tensorflow and sklearn. We also used a number of processing libraries such as pandas and numpy.
Challenges we ran into
There were many challenges we ran into and one of them was optimizing our model to get better prediction results. From the results of the leader board we saw that the prediction accuracy were all pretty high and the difference between the teams were minor, around 0.1%. Therefore we really spent a lot of time fine tuning the hyper parameters of our models, hoping to improve our accuracy.
Accomplishments that we're proud of
We were proud that we make progress in tuning our model parameters and increasing our prediction accuracy. As a result we were ranked top 8 of the leader board
What we learned
We learned more about data Science, machine learning and how to fine tune the hyper parameters to achieve better results.
What's next for ConocoPhilipsChallenge(hardware failure prediction)
Next we hope to explore better models for our data and keep on improving the prediction accuracy.
Log in or sign up for Devpost to join the conversation.