Robot safety part 6: data mining
John Kersey continues his exploration of the methods and techniques used to make AI such a potentially effective tool for safety.
‘The possibilities for creation and insight are endless. We’re constantly collecting more data, and it’s starting to be very relevant to our lives.’
Finding the story in your data
Data mining, when associated with the related activity of analytics, is often called finding the story in your data. Some see data mining as a step on from the discipline of business intelligence which seeks to answer set questions or queries, whereas data mining goes further and looks for hidden factors or patterns at work. This is a necessary pre-cursor for studying these in more detail using analytics and machine learning. Using these related disciplines large amounts of data or Big Data can be processed using techniques such as clustering (how the data is related) and a variety of insights uncovered. These patterns and factors then become part of the narrative and so the analogy with storytelling.
For example a telematics company blog looks at vehicle accidents Case 3. Speed prior to Impact. Although there are many accidents at high speed, as you might suspect the data reveals the majority of accidents (51%) occur below 40mph in city conditions. Even in isolation this might be a surprising conclusion to some but it is the beginning. We want to know more; why this is so and what are the other factors involved (traffic and weather conditions, time of day, type of vehicle and so on) to learn the full story, relay it credibly and gain insight.
9 Laws of Data Mining
When you preparing data and refining the outcome after mining to find the nugget (the insight) then guidance is needed to make sure the effort is in the right direction. Fortunately we have a skilled guide to help us. Tom Khabaza was an early pioneer of data mining back in the 1990’s and derived 9 laws that are still used heavily to review data and are worth looking at in full:
First law. Business objectives are the origin of every data mining solution: If you don’t know what problem you’re trying to solve, you probably won’t solve it.
Second law. Business knowledge is central to every step of the data mining process: If you don’t have someone who knows the business on the team, you won’t get good results.
Third law. Data preparation is more than half of every data mining process: Analytics isn’t always pretty. Most of the time and effort goes into the dirty work of cleaning data and getting it in shape for analysis.
Fourth law. The right model for a given application can only be discovered by experiment: In business applications, it takes a lot of trial and error to find predictive methods that work for you. (This is different from classic scientific research processes.)
Fifth law. There are always patterns: In practice, your data always holds useful information to support decision-making and action.
Sixth law. Data mining amplifies perception in the business domain: Do the analysis and you’ll know and understand more than you did before.
Seventh law. Prediction increases information locally by generalisation: Good analytics processes provide useful predictions and a better understanding of what’s likely to happen in specific business situations.
Eighth law. The value of data mining results is not determined by the accuracy or stability of predictive models: Judge results by the value they yield for the business, not by the mathematical details.
Ninth law. All patterns are subject to change: What works today may not work tomorrow. You’ve got to keep investigating.
(as quoted by Meta S. Brown – author of Data Mining for Dummies).
The implications for safety
Safety is often about changing behaviour and this most often comes from changing our beliefs about a condition or situation. It is often held that people respond best to stories and the art of storytelling has become a much prized skill. How much better if we can use facts and figures (data), engage people with hidden valid information that defines insight. How much better it would be if policies and procedures were formulated on this insight rather than received or perceived wisdom.
To read the other instalments of this series, click here.
Disclaimer. The views expressed in this article are those of the author and do not necessarily represent those of any commercial, academic or professional institution I am associated with.
John will be speaking in two seminars on Tuesday 19 June at the Safety & Health Expo.
AI revolution in safety – Robot safety rocks – shaping the future!
Operational Excellence Theatre, Tuesday 19 June, 14:45.
John will discuss how AI revolution is happening and it is already affecting the safety industry. Will it be innovation or disruption? How will this affect us as organisations, service providers and practitioners? John Kersey, SHP “Robot Safety” contributor will take you through 5 steps to get you started on the Industry 4.0 path.
New safety debate
Professional Development Theatre, Tuesday 19 June, 15:50.
John will part of a panel that provides an introduction to the ‘new safety’ campaign, developed by SHP and the health and safety community. The debate will include results of a survey and will debate the following statements:
Health and safety should be:
1. Holistic – caring for both mental and physical health.
2. Disruptive – embracing new approaches and technologies.
3. Business leadership – driving cultural change from the boardroom to the frontline.
4. Innovative – finding creative solutions to problems.
5. Inclusive – celebrates diversity of views and individuals.
6. Sustainable – champions environmentally sustainable business values.
State of the health and safety profession
Dive in and explore the 'State of the health and safety profession, sourcing PPE in the age of COVID-19' webinar on-demand alongside Andrew Sharman, SHP editor Ian Hart and their guests.
Listen for free today.