search

dancyfang / OR4741_RealEstate_Manhattan

Predicting RealEstate value in Manhattan
0 stars 0 forks source link

Final Peer Review #14

Open zzwustc opened 6 years ago

zzwustc commented 6 years ago

The goal of this project is predicting market value of real estates in Manhattan, New York, which is a meaningful and popular issue. The research is impressive, there are lots of aspects I like about this project:

  1. Convincing graphs
  2. Well applied models taught in class
  3. Training error and test error on the same magnitude, implying no overfitting
  4. Sparse effect is significant with l1 regularizer

Maybe, something could be improved:

  1. Only 500 samples in dataset
  2. low correlation between y and a single feature doesn't imply this feature is useless in the model including all the features.
  3. Personally, I guess outliers cleaning should involve standard deviation
dancyfang commented 6 years ago

Hi! Thank you for your review. You might have some misunderstanding on our report. Our data set includes 42000 original rows each with 84 features, and around 16000 samples after cleaning.