/Click-Through-Rate-Prediction-1

Predict whether a mobile ad will be clicked

Primary LanguageScala

Click-Through-Rate-Prediction

In online advertising, click-through rate (CTR) is a very important metric for evaluating ad performance. As a result, click prediction systems are essential and widely used for sponsored search and real-time bidding.

Feature Selection

  • label : 0/1 for non-click/click
  • id
  • hour => 0600-13:1, 1400-21:2, 2200-05:3
  • banner_pos
  • site_id
  • site_category
  • app_id
  • app_category
  • device_id
  • device_ip
  • device_model
  • device_type
  • C1, C14-C21 -- anonymized categorical variables

Dependencies

  • Spark 1.6.1
  • Scala 2.10.4
  • SBT 0.13.8

說明

總共有三種Model: SVM, Logistic Regression, Random Forest。

Step 1

使用 spark 實作 Recursive Feature Elimination, 找出適當的屬性

Step 2 建模

Hyperparameter tunning: return model which has the best AUC Area

Step 3 投票

每個模型的建立過程會先做以投票方式產生Label。

Step 4 計算投票的AUC Area