top of page

Buzzword Trigger Warning: Applied Machine Learning

timothyjwhite20

"Hesitation increases in relation to risk in equal proportion to age." - Earnest Hemingway


Machine learning has become a favorite term of snake oil salesman in every heavy industry facet. While a decision maker shouldn't necessarily be an expert in this emerging field, they must be familiar enough with it to separate fact from fiction and lost revenue. The quote in the post heading is an admonition I interpret to mean that hesitation when jumping into the deep end of data analytics is just a reaction to potential risk and reward. Machine learning is where we begin to doggy paddle.


Full transparency, in college, I was not the best coder; in fact, I was barely functional. I scraped by an elective, introduction to C++ course with a low C. This course left a sour aftertaste because coding was packaged as more algorithmic and less project-focused or creative. Post-college, I found that this was not the case for most applications. A great resource for beginning to understand the power and fundamentals of machine learning is Udemy.


An intermediate understanding of statistics is preferred, but not necessary. The course is based in the Python coding language, and all the code and datasets are made available to tinker and modify. What is necessary is a willingness to put aside preconceived notion about how difficult a topic might be to learn and apply.


Different methods of regression with explanations, sample code, rationale and pro tips are all discussed in as much detail as the user can handle. What I appreciate is a focus on both the math and the coding to give a fuller picture of capabilities. Applied to my work in mining and reliability, I used this machine learning tool with help from the ML course to create a digital, multi-variate model of a double rotator ball mill, an impressive, huge, and extremely expensive piece of machinery. When test data was fed into the model, it was 89% effective at predicting optimum or sub-optimum output vs. inputs.


While this project stopped when I left the company, I learned the much more valuable skill of separating fact from corporate propaganda. I have had the opportunity to give advice on to superiors on whether a potential software platform is offering legitimate potential improvements or just quoting industry buzzwords. While I will never be the "only" in the worlds of IoT and Machine Learning, I have a new skillset to apply, and it started with 10 dollar course offered online and taken on-demand.



13 views2 comments

Recent Posts

See All

It's Ok To Be Tired

"The only thing that's stressful about being an entrepreneur is every single second of every single day" - Gary Vaynerchuk While I am not...

The Case for Pivoting

Proponents for specialization are everywhere today, it would seem. From LinkedIn to Indeed to Monster, hiring sites and social media...

The Switch() Function in Microsoft Power BI

For those new to programming, an IF-THEN statement is a function that executes a specific code segment if when a certain condition is...

2 Comments


Yussef Alshalaldeh
Yussef Alshalaldeh
Jun 26, 2022

Well said Tim, Machine Learning is the newest Buzz word and many see it as an "easy button". The truth is, many industries are no where near ready to begin applying this level of technology in their business. So many industries have no data or bad data and to get to the point where you can execute on a true machine learning platform takes a lot of diligence and maturity around Data integrity.

Like
timothyjwhite20
Jun 29, 2022
Replying to

Agreed. Even in fixed process plants that are diligent about instrumentation upkeep, cleaning the particular dataset is a substantial task by itself. Luckily, machine learning is evolving to where algorithms can be used to clean datasets for other algorithms. I see no future problems where machines begin learning from other machines and welcome our new robot overlords. https://dzone.com/articles/using-machine-learning-to-automate-data-cleansing#:~:text=The%20first%20step%20where%20machine%20learning%20plays%20a,fit%20in%20with%20other%20values%20of%20that%20column.

Like
bottom of page