Quality Alchemist (品質煉金術師): Minitab Seminar on Gearing Up for Industry 4.0 - Machine Learning for Manufacturing Performance

2018年12月7日星期五

Minitab Seminar on Gearing Up for Industry 4.0 - Machine Learning for Manufacturing Performance

The seminar named “Gearing Up for Industry 4.0: Machine Learning for Manufacturing Performance” was organized by Minitab on 7 Dec 2018, which was one of activities after the 14th Hong Kong Quality Management Convention.

Before the seminar, I took a photo with Minitab team and speaker.

(Left: Mr. Alex Kwan (Country Account Manager), Mr. Darren Chai (GM, Hangzhou Qiyue Data Technology Co., Ltd.), Mr. Calvin Yim (Product Manager, TechMax), Ms. Iza Ng (Business Development Manager, TechMax) and I)

Ms. Angel Wong (Senior Consultant, Electronics Testing and Reliability, HKPC) also joined the seminar.

Mr. Darren Chai (柴磊) (GM, Hangzhou Qiyue Data Technology Co., Ltd.) was guest speaker and his topic separated into two parts. Part 1 was foundation of machine learning in the industry and Part 2 was demonstration of the Salford Predictive Modeler (SPM).

Firstly, Mr. Chai said industry needed to overcome the barrier of data such as leak of capability for data science, suspect the effect after implementation of data technology and traditional IT data isolation. Then he introduced the industry IoT potential as well as data importance.

And then Mr. Chai briefed the trend following industry 4.0 below:

- Digialization in operation (e.g. Smart factory)

- On demand model such as Enterprise as a Service (EaaS) and Manufacturing as a Service (MaaS)

- Partnership in Technology Ecosystem

- Enterprise Integrated (e.g. Vertical Integration)

- Transformation of IT facilities

- Data Scientist Resource

- Automation and Value Chain

- Equipment Maintenance Traceability

After that Mr. Chai briefed Machine Learning and different types of application. Salford Predictive Modeler (SPM) was positioning in the middle level for smart decision.

Mr. Chai also explained the different between traditional statistics and machine learning (ML). Key differences were hypothesis and model that ML did not required. ML was good for solving ultra-high dimensions (multi-variants), complex non-linear model and analyzed human behavior, as well as, complex physical & chemical phenomena.

ML applied in different industries included Bank, Internet Financial Enterprise, Insurance, Telecom, Manufacturing, Sale, Biotech and Internet Company, etc.

There are three key elements for enterprise employing Machine Learning (ML) and they are “Suitable Business Purpose and Scenario”, “History Data for Modeling” and “ML Development Staff and Technology”.

Lastly, Mr. Chai demonstrated Salford Predictive Modeler (SPM) for real situation data set. Old customer survey for 1^st generation of mobile was used.

Selected data set and built the model. Target variable was “Response” (either 0 or 1 / -1 or 1) and then select Predictors.

Classification Tree Topology for Response was generated.

Each node ratio and important ranking was calculated automatically.

The relative importance of each variable was calculated.

At the end, Mr. Chai demonstrated the industry data which had more than thousands of variables that could be run for 2 or 3 time so as to reduce the variables after remove the non-relevant variables in each run. It could also check the relation between different variables in 3D diagram.