Group H proposal
Title: Ali Taobao Ad Recommendation System
Topic Description
In our project, we decide to Predict the target users for Ad propagation by using recommendation algorithms – Item-based Correlation Filtering. The data we use is provided by Tianchi Big Data Competition, which covers 1.1 million users 8 days clicking behaviors under Ad propagation, and Ad & User basic information. How to recommend a specific Ad to potential target user groups for better users’ response is a question remains to be solved. Therefore, we propose this project to use recommendation algorithm to find the potential target user groups for advertisement.
Methodology
Since the entire size of the dataset from Tianchi official dataset is near to 4GB, we decide to use Hive for data ETL and Hadoop for distributed computing to reduce computing time expense. On the consideration of algorithm of recommending ad, we proposed to use Item-based Correlation Filtering (CF) which has been proved to be useful in E-Commerce item recommendation. Specifically, we plan to get user-commodity preference matrix first and using Clustering algorithm to find the similar ad. The recommended ad will be mapped to users who have similar preference on commodities.
Demo Plan
The demo will be a web app recommendation system. User can input an ad info:
Input
target commodity name:
target commodity brand:
target commodity price:
target target commodity brand:
Then, recommended user info will be shown on web page with forcasting Click-through Rate (CTR) possibililty.
Output
Ad is recommended to user with userID1
Ad is recommended to user with userID2
Ad is recommended to user with userID3
Ad is recommended to ...