Differences

This shows you the differences between two versions of the page.

Link to this comparison view

cgss15:students:jiayil:polyanalyst:weeklymeetings:3_25_2016 [2016/03/25 16:28]
jiayil created
cgss15:students:jiayil:polyanalyst:weeklymeetings:3_25_2016 [2016/03/25 18:14] (current)
jiayil
Line 1: Line 1:
 +====== Megaputer Intelligence Weekly Meeting 3/25/2016 ======
  
 +20 minutes for each speaker at Eli Lilly, plus 10 minutes Q&​A. ​
 +
 +Yilong: scheduled to do practice talk next Friday, Jeff also on that day.
 +
 +PolyAnalyst 7 loads data 20 times faster than PA6. Start designing the interface of PA7.  ​
 +
 +More data from Select Medical, have 3 or 4 people work on it. Results should be ready by next week. 
 +
 +Ask people that are to be presenting at Eli Lilly to send Sergei a paragraph or summary. ​
 +
 +PA can use Facebook and Twitter API to load data. 
 +
 +Content providers can change rules or formats at any moment for their literature. See if can make Lilly interested in this project, if so then need one person to work on this project full time. Similar system from Israel. ​
 +
 +===== Eli Lilly Practice Talk - MAUDE Analysis by Junjie =====
 +
 +==== Text Analytics: Medical Event Adverse Event Reports ====
 +
 +Motivation:
 +  - device manufacturers
 +  - FDA, etc
 +  - user review
 +
 +Case Study (analysis of ~80,000 records): ​
 +  - difficulty of manual way
 +  - how easier it is for using text analytics
 +  - what else can text analytics bring
 +
 +MAUDE: hosts reports of adverse events involving medical devices
 +
 +Data cleansing: misspelled words, duplicated entries, empty data, discrepancies,​ etc. 
 +
 +NOTE - Pay Attention to Such Format When Presenting: ​
 +
 +Data Cleansing: ​
 +  - misspells
 +  - duplicates
 +  - empty data
 +  - discrepancy
 +
 +Identify issues: combina data (keyword extraction, auto-taxonomy),​ categorize issues based on taxonomy. ​
 +
 +Analysis & visualization:​ different pie charts. ​
 +
 +Any trends in the discovered issues? Top 5 issues for Roche, Abbott, Lifescan based on their medical device (glucose meters, specifically) reports. ​
 +
 +How good is the taxonomy, how large is it, how many categories, how to make sure nothing is missed, how many records are uncategorized. ​
 +
 +Showing example: don't highlight multiple things at a time, just highlight 1 and make a point about it. 
 +
 +When it comes to the conclusion, the fewer words, the better. Also, don't need the "​."​ at the end of each sentence. ​
 +
 +Try analytical client/web report client and see which way will generate better figures.  ​
cgss15/students/jiayil/polyanalyst/weeklymeetings/3_25_2016.txt ยท Last modified: 2016/03/25 18:14 by jiayil
 
Except where otherwise noted, content on this wiki is licensed under the following license: CC Attribution-Share Alike 4.0 International
Recent changes RSS feed Donate Powered by PHP Valid XHTML 1.0 Valid CSS Driven by DokuWiki