AaronZang / ORIE4741-Home-Purchase-Assistance

0 stars 0 forks source link

peer review #3

Open alexisrc1 opened 8 years ago

alexisrc1 commented 8 years ago

The idea to predict the fair price of a house given previous data, the location, the year the house was built... is a very good project idea. The graphs you show in your reports are interesting but obviously you haven't been far enough in the data cleaning as many houses prices are close to 0. I do not know where your data come from but there seems to be a lot of outliers/missing data. The auto-regressive plot is a good idea and we see a break in 2008 but how can we explain that it only happens for Manhattan? Also, the prices in Manhattan are 4 times bigger than prices in the other districts, it seems a bit weird. I also checked your notebook and the average price in Soho is over 3 million dollars! It seems quite unaffordable... I think that the next plots you show in your report are interesting, even though there should be more colors ! It is appreciable that you tried to use regression that we learnt in class, but your results are obviously way off, I am sure you will correct it for the final report :) I understand that you want to include new features in your model but you should definitely review your current model before doing so as it is obvious (given the results you got using linear regression) that the current features you are using predict very badly apartments prices

Good luck !