cshjin / cs412yelp Goto Github PK

• clean the format
  • string data into cataglory int format
  • None or NaN data into -1
• Convert reviews into a list of dictionaries
• Apply sklearn function DictVectorizer

• test sample dataset format 1
  csv format with user busi csv
• test sample dataset format 2
  json format

{user1:
{busi1:1, busi2:5},
user2:
{busi1:3, busi2:4},
...
}

• based on user's review
• based on user's profile

use SelectKBest function to select 20 best features