Midterm Report Review: zw243 #10
Description
ORIE4741-Airbnb-project
Midterm Report Review
Reviewer: Zilong Wang (zw243)
Abstract + Intro?
The first 2 sections should be swapped, as the goal is the motivation of the
project, and the description and context usually comes next.
Due to the large amount of features, it is understandable that the authors
cannot exhaustively list the features easily, however, a few examples should
suffice: e.g.
Has Wifi? -> Boolean
Review Scores? -> Float
URLs -> Nominal
Policies -> String
etc.
Cleaning of Data
Filtering out entries that do not have a review score seems like a reasonable
approach, as our intuition tells us that the review score should be an
extremely important predictor.
However, as ~ 25% of the data has been deleted, it seems somewhat worrying at
first, though the sheer scale of the remaining data (28071 out of 38810)
mitigates this reasonably well.
Visualizations
The plots are chosen very well, and some make immediate sense.
For example, I would expect more people to take temporary abode in the more
populous boroughs such as Manhattan and Brooklyn as compared to the Bronx.
The bar charts also show a clear trend in the increase in price along with
the increase in number of rooms/amenities, and provide a quick (and simple) way
to confirm the reader's hunch that it should be so.
Regression Analysis and Future Methodology
As this report was probably written before a more diverse class of loss
functions and feature transforms were introduced, the authors themselves
should know what other possible methods to use in regression.
I like the ideas to avoid underfitting/overfitting, some of them are trivially
obvious, such as removing outliers (unreasonable prices) and reducing the number
of features, but grabbing additional reference data, such as home prices, sound
like a good way to sanity check the data.
Overall Comments:
A very well thought out and written report. Analysis done is simple to
understand and conclusions were straight to the point.