Upload
ngonpham
View
2.176
Download
2
Embed Size (px)
DESCRIPTION
Big Data for Startups - Practical tips for applying Big Data for your own startups
Citation preview
Big Data for StartupsNgon Pham - Lead Engineer & CTO at Lana
Content● Introduction● 10 practical problems & solutions● Q & A
Introduction● How to make use of social platforms to
empower your startups○ Facebook○ Youtube○ Foursquare
Problem 1● How to integrate Facebook user information
into your product● Solution
○ Allow Facebook sign-in, ask for related permissions○ Use the token to build user reflection for your
product: Bio, Statuses, Events, Groups, Likes…○ Demo
Problem 2● How to integrate Facebook into your vertical
eCommerce site for social marketing● Solution
○ Integrate Facebook sign in, ask for token to analyze user bio, their likes and compute user interest score ~ how they are interested in your category
○ Encourage them to like your page, share your product basing on their interest score
Problem 3● How to analyze Facebook users, and do the
target marketing even if you don’t know how to code
● Solution○ Use http://www.ocommunity.net/○ Facebook ad targeting by user ID○ Demo
Problem 4● How to crawl all Foursquare places for a
specific country● Solution
○ Use a simple spreading algorithm○ Demo: Vietnam, Singapore, Indonesia, Cambodia
Problem 5● How to use Foursquare as a social layer to
empower your reviewing site: Travel, Food...● Solution
○ Do a matching (manually / semi-auto) between Foursquare places and your database
○ Constantly get updates from Foursquare for: Number of checkins, Photos, Tips...
Problem 6● How to crawl all videos for a specific
language on Youtube● Solution
○ Use a language detection library https://code.google.com/p/language-detection/
○ Crawl using Related Video API provided by Youtube○ Demo
● How to build a classified video app on top of Youtube
● Solution○ Crawl all videos & develop some filtering algorithms○ Some useful tools for cleaning / sampling data:
Google Refine, R…○ Demo
Problem 7
● How to scale processes across machines?● Solution
○ Use a Queue server for dividing the workload https://www.rabbitmq.com/
Problem 8
● How to export processed data for end-users as a website
● Solution○ Use Wordpress
■ Data model is very simple ~12 tables■ Tons of plugins and themes
○ Demo
Problem 9
● How to export processed data for Business Intelligence / Visualization
● Solution○ Export processed data to MySQL○ Use Pentaho / JasperReport to hook in○ Demo
Problem 10
Q & A
Q & A