Inspiration

Some businesses need to analyze their data but they do not have opportunities to hire a data analytics department. Moreover, there are many routines in data science that could be and should be automated. Also, there is a problem of a validation of a trained model, which is also very important problem.

What it does

Performs a training and evaluation of a model on a given dataset in a automated manner.

How I built it

We build prototype in jupyter notebook. We are going to deploy solution with a docker.

Challenges I ran into

Train stage includes many steps that just need to be written. However evaluation stage can be a little bit challenging in case when we do not have a dataset for evaluating it. So we need to do some calculus basing on more generic things.

Accomplishments that I'm proud of

What I learned

What's next for automl

Build automated deploy with re-training and re-evaluation.

how to evaluate

unzip result.zip from google drive install all the requirements (not listed) it is compatible with python 3.6.6 build your model according with the examples dump model to the models dir ... profit

Built With

• fasttext
• lightgbm
• python