What is the problem you are trying to solve?
Satellites around the world are constantly taking readings using their sensors. Some are in search for a pulsar star. However, correctly identifying one out of all the readings takes a lot of efforts and what’s worse is, it takes a lot of time. However, thanks to machine learning and dataset available through the hard work of astronomers, it is possible to predict the existence of a pulsar star which can later be confirmed by the astronomers.
Briefly explain the models you experimented with to solve the problem.
I tried various models such as Logistic Regression, K-Nearest Neighbours, Random Forest Classifier, Support Vector Machine and Decision Tree Classifier. Not only did I try training them using their default values (using sklearn), I also tuned the hyperparameters to conclude the best model to use for this problem.
What is the result you are the most satisfied with (accuracy, other metrics, etc)?
After training the data using the various models and hyperparameter tuning the models, I tested the models on a validation set I had kept aside. Based on both the accuracy of the models and confusion matrix, I concluded that Support Vector Machine outperformed all the other models.
If you were to continue this project, what would you explore doing/improve?
To begin with, I would have stricter data input validations. The webapp currently crashes when an invalid data is input. Next, I would include the option to upload a csv file with correct formatting and feature headers, to batch process and make predictions on the provided data. I would then return a table containing the positive results.
Log in or sign up for Devpost to join the conversation.