Python for Data Analysis - Project
David Azoulay & Inès Nguyen DIA1
Avila Dataset
To do the final prediction, we did the following steps on jupyter notebook:
- data visualization & exploration of the training dataset
- choice of the best model trying different ones, using cross-validation and grid search to boost the parameters
- final prediction on the testing set
We obtained an accuracy equal to: 99.6%
You can find our results is the file prediction_responseVSactual_classes.txt which is in the folder named avila
We did an API which asks to a user to enter a value for several attributes (not all of them to avoid you losing your time ;)). Then, we use the model we found before to make a prediction on the class it belongs. The API is made with Flask and uses a .py file and .html file (as a template). To be sure that the API works well, make sure that you met all the requirements and then go on http://localhost:5000/
All the steps we followed, how we proceeded and the results we obtained are described in the report.