User analysis in line today project
2 likes507 views
This document provides an agenda for a user analysis presentation on LINE TODAY, a news service in Taiwan, Thailand, and Indonesia. It discusses why user analysis is important to understand user characteristics and enhance user experience. It describes how LINE TODAY conducts user analysis using Spark to process log data from ElasticSearch and Kibana for visualization. A case study example evaluates the performance of a LINE Points event and analyzes user demographics and behaviors. Tips are provided on understanding Spark execution plans and using functions for aggregation, as well as tips for setting up mappings and visualizations in ElasticSearch and Kibana.
User analysis in line today project
- 2. USER ANALYSIS in LINE TODAY Yang Yachu (Rita), Taiwan Engineering
- 3. - About LINE TODAY - Why User Analysis - How to Do User Analysis - Case Study - Tips You Should Know Agenda
- 4. LINE TODAY A News Service in Taiwan, Thailand, and Indonesia
- 5. - Understand User Characteristics Why User Analysis - Enhance User Experience - Evaluate Marketing Event Performance
- 6. How to Do User Analysis Architecture Log Server Analysis (Spark) Visualize (ElasticSearch Kibana) Line Today frontend
- 7. Spark Source and Computation Actions (Line Today) - Read - Like - Comment Demographic Prediction (Datalab) - Gender - Age - Region Demographic Data Action Raw Log Daily Action Data Count, Average Action x Demographic
- 8. Elastic Search + Kibana Visualisation
- 9. Case Study Evaluate the performance of Line Points Event - We would like to know the overall behaviour changes
- 10. Case Study Evaluate the performance of Line Points Event - What kind of users are interested in Points Event? Days Distribution Age Group
- 11. - Understand your execution plan: df.rdd.toDebugString, df.explain Tips - Lots of useful function for aggregation: spark.sql.functions About Spark
- 12. About Spark Understand execution plan Transform and Action - Spark is lazy - Action triggers transformation
- 13. - Setup mapping: date is important, text and keyword behave very differently Tips - Date Histogram: change aggregation interval, don’t do it by yourself - Scripted Field: project new filed based on existing field About ElasticSearch and Kibana
- 14. Thank you