#UEF

• data
  • http://data.computational-advertising.org/
  • ftp://140.113.213.5/02_LAB/2_畢業資料/碩士/2016/王俊儫/data 這個是我從上面的連結抓下來放的，原本的下載很慢
• dependent libraries
  • pymongo, scikit-learn

##Library

• lib
  • classifier: CTR prediction models
  • featureDistribution: Calculate conditional mean for each hashed feature
  • featureHasher: A streaming feature hasher dealing with nested dict based on sklearn FeatureHasher
  • featureSelector
    • poissonInclusion.py
    • significantFeatureSelector.py: select feature using featureDistribution
  • fileDB: read data

##Experiment

• lr: LR
  • offlineMemory, offlineTime, onlineMemory, onlineTime, standard
• lr_fh: LR_FH
  • offlineMemory, offlineTime, onlineMemory, onlineTime, standard
• ftrlProximal: FTRL-Proximal
  • onlineMemory, onlineTime, standard
• sem: UEF_SEM
  • offlineMemory, offlineTime, onlineMemory, onlineTime, standard
• ssem: UEF_SSEM
  • offlineMemory, offlineTime, onlineMemory, onlineTime, standard

# standord: evaluate AUC and Loglikelihood for the model

##Plot Data

• plot
  • e1: Stability of Index Set
  • e2_e3: Select size k vs AUC, loglikelihood
  • e4: Online Memory Usage
  • e5: Index Set Size k vs Offline Memory Usage
  • e6: Online Run Time
  • e7: Index Set Size k vs Offline Run Time
  • e8_e9: AUC, Loglikelihood
  • e10: Distingusiablity of Each Feature
  • e11: Index Set Size k vs Online Memory Usage
  • e12: Offline Memory Usage
  • e13: Index Set Size k vs Online Run Time
  • e14: Offline Run Time