nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

12/01/15 - JMP EBC Internal Planning Visit Call Notes - GE Power - Atlanta, GA

scwise

Team,

Notes from our JMP EBC Internal Planning Visit Call Notes –

12/01/015 - GE Power - Atlanta, GA.

JMP EBC Internal Planning Visit Call Notes – 12/01/015 - GE Power - Atlanta, GA

1. Place, Date & Time

Customer: GE Power
Place: Atlanta, GA
Dates: 12/01/15
Times: 8AM-9AM CST
Meeting Type: JMP Internal Planning Visit Call Notes

2. Background

Meeting to discuss setting an agenda for the GE Power EBC Visit scheduled for Dec 14th & 15th
- Output from 11/10/15 Site Visit to Sameer’s Fleet Management Analytics Team in Atlanta, GA (Notes: https://community.jmp.com/docs/DOC-9454)
Background on Sameer Vittal
- Sameer is Manager of Fleet Management Analytics in GE Power in Atlanta
- Linked-In (https://www.linkedin.com/profile/view?id=AAEAAABK8i4BaAgCIGok5oxU_Q0JWPf2JD2BlXY&authType=name&authToken=mkbx&trk=prof-sb-browse_map-name)
- Former SAS champion, now moving to JMP
- Not the Decision Maker, but influential and very close to the top levels of Management
  - Reports up to Justin Eggart, GM Engineering (https://www.linkedin.com/in/justin-eggart-3480a788)
- Has team of Data Scientists and Analyst that supports Turbine Plants Services out of Fleet Management
Sameer’s Fleet Management Team:
- Provides Service Offerings for:
  - Power Plants using GE Turbine Products
  - Power Plants using other Company Turbines (Siemens, etc.)
  - Branching out into other areas such as Natural Gas
- Works with the Product Group in Atlanta which is a new audience for us
- Also Team is working with French power acquisition of Alstom (big potential for us) (http://www.ge-alstom.com)
Current Analytic Tools Used:
- Using Predix: GE Digital Platform for Energy, Healthcare, Transpiration & Aviation (http://www.ge-alstom.com)
  - Cloud Based, Includes Analytics
  - "The Predix Platform-as-a-Service (PaaS) helps you develop apps on cloud infrastructure that connect industrial machines with people through data and analytics for better business outcomes”
- Use R in the past and now moving to Python based for data management, acquisition and model automation
- JMP Pro modeling rewritten in Python for uploading
- Additional Info on data machine connection: Internet of Things (http://www.ge.com/stories/industrial-internet)
Overall JMP Opportunity:
- Tremendous upside for growth of JMP for the turbine business
- Can help Partner with Development in new and upcoming areas
- Fifty JMP Pro possible through Sameer

3. Attendees:

JMP:
- Jon Weisz – VP, JMP Sales & Marketing
- Dan Valente – JMP Program Manager
- Dan Durkin – JMP Global Account Manager
- Scott Wise – JMP Principal Systems Engineer & Technical Global Team Lead

4. Meeting Purpose

Discuss setting an agenda for the GE Power EBC Visit scheduled for Dec 14th & 15th

5. Customer Needs

Sameer: From what I’ve seen so far .. these are new features that will be very helpful for us in Fleet Management ..and for engineering / industrial analytics in general. The use cases are anomaly detection (time series), topic identification (text ), data-mining and the usual focus on quality, reliability, etc. Gaps we see in JMP now are mainly in feature creation, where signal processing is useful.
- Text Mining
  - Pre-processing (stemming, roots, whitespace removal, case conversion, using custom-defined stop word lists, having a way of selecting multiple stop words from the word-count)
  - Ability to identify top issues (topics) using supervised and unsupervised learning, so we can classify warranty issues into the right failure mode and component
  - Linking text mining scores to the rest of JMP/Pro’s data mining capabilities
- Regression & Data Mining
  - Random forest tuning (similar to capabilities available in the caret package in R, that we use extensively)
  - Generalized Additive Models
  - Naïve Bayesian classifiers
  - More mixture modeling capability (especially Gaussian mixture modes, E.g. equivalent to the mclust package in R)
  - Mixed Effects Logistic Regression (this is very useful)
  - Clustering categorical data (K-Modes and K-Mediods)
  - Methods for correcting for class imbalance (this is very common in our datasets)
  - More in association rule mining (should include temporal ARM as well as standard, as we look for rules from events like patterns of alarms detected in time)
- Multivariate
  - More focus on outlier detection (E.g. using local outlier factors) in addition to all the cool stuff that’s being put in there. We think of it in the following hierarchy
  - Distance based outliers (T^2, Mahalanobis, Local Outlier factor)
  - Density-Based (this is where mixture modeling comes in, again)
- Time Series
- New areas in the future
  - network analysis, spatial data (builds on mapping capability that’s available .. especially things like point pattern analysis, spatial autocorrelation, etc..
  - Spatial analysis could be important going forward, people are just getting started on using it for survival and risk analysis for industrial assets
- Additional Notes:
  - All of this is in open source and very popular in the industrial analytics community.
  - There is a new textbook called “Data Mining : The textbook” by Charu Agarwal (Springer Publications, 2015) who is a well-known researcher, part of IBM research.
  - We use his book a lot as a guide to the various methods used in industry, it’s not specific to any language and provides a good theoretical overview of the methods I have mentioned above.

6. What Discussed

Participants:
- From GE:
- From JMP:
  - Jon Weisz
  - Dan Valente
  - Dan Durkin
  - Scott Wise
  - John Sall & Select Development (Invite & John will also join for lunch)
  - Leo Wright (Invite for possible System Reliablity Info)
Better understand how they are using JMP
- Concern with using standard pricing if this is part of a service offering to other customers
  - Two ways we get funded:
    - Funding it through bigger volume standard pricing (benefits only to GE)
    - Fund it through using select JMP incorporated in a service offering
- Make sure we understand why they use what tools and listen carefully
  - Dan’s old notes: Wants to run 5 Random Forest Models and save out to SQL. In the process he has to define “interconnects” and “output variable” which he had not found a way to do in JMP, or at least not easily.
  - Consumers of the NPI Team’s System Reliablity data
Sameer’s Goals:
- Wants to understand where JMP is going in relation to the topics he requested (Also see Customer Needs)
  - Text Mining
  - Regression & Data Mining (Advanced Models)
  - Multivariate Anlaysis (Outlier Detection)
  - Time Series Data Mining (Signal Detection)
  - If they are serious, they can simulate data without sending real data
Python Discussion
- Ask question: If doing everything in Python, why JMP
- Python using for streaming data
- Dan’s follow-up info on Python
  - SAS v Python for Data Analysis: http://www.sasanalysis.com/2014/03/sas-vs-python-for-data-analysis.html?utm_source=feedburner&utm_medium=feed&utm_campaign=Feed%3A+SasAnalysis+%28SAS+ANALYSIS%29
  - Python and Hadoop: http://blog.cloudera.com/blog/2013/01/a-guide-to-python-frameworks-for-hadoop/comment-page-1/#comment-34718
  - Python Stack for data analysis: http://pydata.org/downloads/
Agenda review:
- Late start on Monday, Dec 14th (10-11AM) due to Sameer’s travel in Monday morning
- Bulk of topics Monday afternoon and Tuesday morning
- Open/Q&A in afternoon of second day so can end the visit earlier as needed
- Good to have John there on lunch and invite him to the whole thing & invite members of his staff
- Dan to talk to John on how want to participate on the agenda
  - Chris G. may be eager to join as well
- Dan’s Proposed Agenda so far (times to be adjusted)
  - Applied Materials Visit to SAS, Institute December 14-15
    - Executive Briefing Center (EBC) Building C
    - Room EBC1 (Or TBD)
  - Draft Agenda: Day 1 December 14
    - 9:00-9:30 Breakfast Buffet and Introductions EBC1
    - 9:30-10:00 Jon Weisz, Dan Durkin: Welcome and Agenda Review
    - 10:00–11:00 Text Mining
    - 11:15-11:30 Break
    - 11:30-12:00 Text Mining con’t
    - 12:00-1:00 Lunch, Executive Dining Room
    - 1:00-2:00 Regression and Data Mining
    - 2:00-2:15 Break
    - 2:15-3:00 Regression and Data Mining con’t
    - 3:00-3:15 Break
    - 3:15-4:30 Multivariate Analysis
    - 4:30-5:00 Recap of Day 1, Topics for Day 2
    - 6:00-9:00 Dinner TBD
  - Day 2 – August 11
    - 9:00-9:30 Breakfast Buffet EBC1, Recap of Day 1
    - 9:30-10:30 Time Series Data Mining
    - 10:30–10:45 Break
    - 10:45-12:00 Time Series Data Mining con’t
    - 12:00-1:00 Lunch
    - 1:00-4:00 Other topics based on timing and review from Day 1

7. Customer Feedback

N/A

8. JMP Opportunity

Dan Durkin’s Goals:
- My goals are to firmly plant JMP Pro as the standard tool for Sameer's team and the wider teams he supports.
- Fifty new JMP Pro users is possible.

9. Next Steps

Dan D. to touch base with Sameer once he is back to US
Dan V. to adjust agenda times
Dan V. to invite Leo Wright for possible System Reliablity tie-ins
Dan V. to invite John Sall and any of his recommended Development Team to attend sections they feel would be beneficial
Scott W. to post notes and relevant info