Four conclusions I made about the dataset:
- The restaurants locations are scattered and there are no evident clusters. However, we can still use K-Means to group them together. For more details, please view the
Location Clustering
notebook. - By using Elastic Net Regression and taking the log of
Votes
andReviews
, I have managed to predict theCook_Time
with a Mean Absolute Error of about 8 minutes. For more details, please view the notebookCook Time Prediction
.- I have also found out that restaurants that serve only Fast Food are quicker in preparing their meals than ones that don't by nearly five minutes but this effect should not be considered significant.
- I have managed to find out a scoring algorithm that takes into account both the
Rating
and the number ofVotes
. For the detailed algorithm, please view theRestaurant Scoring
notebook. - An interesting fact is that coffee is mostly served with Fast Food, other beverages, and Italian food. This conclusion is reached by examining the
Cuisines
column. You can try doing the same thing with other cuisine, too. For the details, please view theEDA
(Exploratory Data Analysis) notebook. There are also other comments I made in this notebook.
For a better view of the notebook, please visit this link: https://nbviewer.jupyter.org/github/dcaohuu18/Xtern_20_DS/tree/master/.