1
$\begingroup$

I am a beginner in data science. I have a data set of drivers that has the following attributes available-

  1. Time stamp
  2. Speed
  3. Acceleration
  4. GPS co-ordinates

I need to build a driver rating system to rate drivers on a scale of 1-100 on parameter of speeding. The data set has the data from many drivers and also their past data. Which is the best algorithm that I can use in this situation?

$\endgroup$
1
$\begingroup$

You need to define what speeding means and how to rate the drivers for amount of such. As you note that you have speed in your dataset this should be rather trivial. Not sure at which stage you want to involve machine learning? Seems like you don't need any. Machine learning is not the same thing as learning about cars and their drivers.

| improve this answer | |
$\endgroup$
0
$\begingroup$

You seem to be trying to do supervised learning without having labeled data.

Do you have a general understanding about Machine Learning?

Usually the ML problems are divided into 2 categories:

  • Supervised Learning: Where you have data samples and a label for each sample, the goal is then to predict what are the labels of new samples.

  • Unsupervised Learning: You just have the data samples and no labels, you want to find some way to group the data or to find some patterns in it.

You should check the scikit-learn tutorials there you can learn the basics about ML and probably learn what you need for your project.

| improve this answer | |
$\endgroup$
  • $\begingroup$ Yes, I had a basic understanding and I know about supervised and unsupervised learning methods in a general sense. I was stumped at this problem because it felt like a Supervised learning problem but there was no labels. Can you shed more light on what algorithms can be used in this case? Thank you for the link to the tutorials, I'd definitely go through them and try to figure this out. $\endgroup$ – Shileen Upadhyay Mar 14 '16 at 17:23
0
$\begingroup$

You need examples if you want to regress ratings. I would get someone with domain knowledge (product owner for example) to take a look at a subset of your data, let them rate those data points and use this subset to train a model on your features and maybe some added artificial features. Then you can use this model to rate the other drivers.

| improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.