Customer Keynote: PIXNET Media Inc.- Business Intelligent and Analysis: Empirical Case Study in PIXNET

© 2016, Amazon Web Services, Inc. or its Affiliates. All rights reserved.

Dr. Rick Lu Head, Research and Engineer, PIXNET 2016.05.20

Using Big Data Mining for Business Intelligent and Analysis in PIXNET

Outline

•  About PIXNET: The dominant BSP in Taiwan •  Impact from “Big Data” Era •  BI Model: Collection, Analysis, and Prediction •  Why AWS ? •  Q&A

About PIXNET

Alexa Ranking #1 (Taiwan) #76 (Global)

Registered Members 5M

Daily Unique Visitors 7.7 M

Mobile TrafGic 52%

Number of Blogs XXX M

Number of Albums XXX M

About PIXNET Services •  Blog Service Platform (BSP) •  PIXinsight (Big Data Service) •  AD Network ( ) •  Business Model •  Advertisement (Banner, Event sites) •  B.I. insight report •  AD Network

Precise x ROI

Impact of Big Data Era Q1: How many articles about “me” in PIXNET ? Q2: Are these articles, Positive or Negative ? Q3: How many visitors regarding on the articles ? Q4: Any profile briefing on the visitors ? Q5: Any marketing strategy providing ? …………………………

社群 會員

MIB AD Sys. Smart Ranking Content Marketing

Big Data Mining

具有會員功能 之社群機制

20162014 ~ 2015Future Plan in PIXNET

Possible System Framework

8Article

Tag User Tag

Access Log

Article Indexing

Keyword Relation

Database

Custom Report

UI

User Tagging Rule Define

UI

User Tagging Engine

Tag Retrieval API

Rule Base Tagging Engine

Report Engine

Article Tagging Engine

廣告系統

商品投遞系統

文章推薦系統 投放規則 設定 UI

營運人員 PM

PIXI

NSI

GH

T

外部電商 推薦系統

Public API

推薦系統

10

11

12

13

Cloud Solution: using AWS

RMDB, S3, EC2: •  Easy-Scaling UP/Down •  Acceptable Pricing

Redshift: (from Oregon to Tokyo) •  millions of blogs & billions user logs •  Easy Analysis for Research Scientist

Summary: Big Data in PIXNET Infrastructure

•  Efficiency, Scalability!! (Thank you very much, Amazon) •  Convenient platform required!!

Text Mining •  Opinion Mining for Positive and Negative Done •  Spam article detection Done, Accuracy 90% (Patent)

Social Analysis •  Spam Authors identification Done, Accuracy 81% (Patent)

Visitors Profile Estimation •  Gender estimation Done, Accuracy 79.2% (Patent) •  Age Done, Accuracy 68.3% (Patent) •  Reading Intent (working) •  …

15