Return to site


Is planning going Sci-Fi?

· AI,Algorithm,Machine Learning

Big data seems to be the magic word of this decade and it becomes more and more important. Provided that it’s applied right and in combination with artificial intelligence, it can give companies a head start in planning, analysis and decision making. It’s about measuring what matters. For every company or project that can mean measuring totally different variables. A complex puzzle that requires quite a sophisticated expertise. At Calendar42 that’s where our data scientist Gürsel Karaçor comes in. In this episode on predictive analytics we start with what it is and what can it do.

Predictive analytics - ©Gartner

Source: Gartner

Forecasting 2.0

During the talk with Gürsel, he explains how self-learning algorithms automate complete scheduling processes, machines and robots. For me as a non-techie it’s actually hard to understand on his level, but one thing becomes really clear to me: there are two areas that seem to be predestinate for artificial intelligence (AI), being comprehensive planning problems and complex control processes (e.g. autonomously driving vehicles). The most appealing example of planning is forecasting.
Forecasting is a sort of predicting. No crystal bowl mystic, but the work of an expert who is watching all sorts of influencing factors and can tell what likable situation you have to deal with. A simple example is whether we need to take an umbrella tomorrow with us or not. Forecasting can be very useful, but predicting supply and demand is difficult and doing it in this way often results in disappointing outcomes, followed by incorrect decisions.
Therefore no persona’s, no ideas, but actual facts. Why?

Rule-based systems mostly fail, because the predictions don’t come from data collected in the past but from assumptions derived from external factors.”, says Gürsel.


It’s clear to me that predictive analytics is a holy grail in solving complex planning issues, but how can it be applied?
Artificial intelligence (AI) is a powerful boost for predictive analytics. It can improve outcomes immensely by accurately modelling the variability in demand. In other words, you will get a much more reliable result when you take a lot of historical data, learn the machine how to deal with them, so it can analyse the likelihood that something will happen.
For this, all relevant inputs (data) are continuously analysed and at the same time, disturbances, such as accidental fluctuations are filtered. The system learns from the data and adjusts the model to changing behaviour. The control algorithms then use these models and combine historical data with real time data. Integrated with the planning and process control systems, they ensure the desired optimisation of logistic processes.
Let me give you a few examples.

Examples predictive analytics

Predictive maintenance is a practical example of predictive analytics. Businesses that rely on many machines and installations have the aim to minimise maintenance costs. 100% availability and reliability of critical installations and machines is crucial for their business continuity. Digitally monitoring these installations and analysing the data will help to perform preventive maintenance tasks and reduce disruption risks.
Blue C.R.U.S.H. of the Memphis Police combines historical data with real-time data to predict what are likely to be the crime hotspots. Now the police knows what locations in the city are the most effective for them to be present in for prevention purposes.
The London Fire brigade uses predictive analytics to predict which properties are at the highest risk of fire. That enables them to carry out targeted inspections to prevent properties from burning down. In order to do this, they use sixty different parameters, including the year of construction, the destination and the number of fire incidents in the vicinity. The predictive software then provides a list of properties that are at higher risk of fire.
Less ‘planning-wise’, but something most of us will recognize is Netflix. Netflix has over 100 million worldwide streaming customers. Having this large user base allows the video streaming company to gather a tremendous amount of data. With this data, Netflix can make better decisions and adjust their user service and offerings on an individual level. They get answers to whether it’s important that credits roll or how much content users need to watch in order to be less likely to cancel.
Since predictive analytics is also really useful in planning and mobility we also apply this to our current projects. Stay tuned, we will share a view of those in the near future!

About the author
Kelly Turenhout is communications manager at Calendar42. She manages our online communication and writes content for our different social media channels.

About Gürsel Karaçor
Gürsel Karaçor is senior data scientist at Calendar42. He builds predictive models for various different customers. He has a PhD. in financial time series prediction.

All Posts

Almost done…

We just sent you an email. Please click the link in the email to confirm your subscription!

OKSubscriptions powered by Strikingly