Machine Learning powered by SAP SCP
Do you want to predict errors, improve your delivery reliability or reduce downtime costs? Would you like to make better use of the data history built up over decades? Do it now - with SAP SCP
Frequently asked questionsSee all questions
Make better decisions automatically
Then you are directly in the subject area of Machine Learning. Machine Learning, Predictive Analysis/Maintenance and Artificial Intelligence - there is currently hardly a trend that is considered in a similar way to machine learning. Gartner's Hypecycle 2016 puts machine learning at the top of the list and predicts that it will be widely used over the next 2-5 years. This development should undoubtedly be seen as one of the most important issues in 2017.
Artificial Intelligence tries to map human intelligence and learning ability in computers.
Artificial Intelligence (AI) as a sub-discipline of computer science tries to reproduce human intelligence in machines and to develop capabilities that go beyond human intelligence through machines.
AI thus forms the umbrella term for all technical attempts to make machines more intelligent. The AI is divided into several sub-areas. These are, for example, Natural Language Processing, i.e. the understanding of human language by machines, Computer Vision, whereby machines are to be taught to be able to interpret images in exactly the same way as humans, but also the subject of Machine Learning (ML).
Machine Learning uses existing data to generate models. If these ML models are used to predict future events, values and information, they are referred to as predictive analytics.
When the prediction of optimal maintenance times is the focus of the ML model, predictive maintenance is often referred to as maintenance that no longer takes place at fixed times or even only when a device or machine has failed.
Three steps to a successful machine learning project
First, a model is trained with existing data. This means that the machine is shown examples containing all the information, including the information to be predicted. These examples are processed and learned by the machine. However, there is no simple memorization, rather the algorithms recognize patterns in the examples provided.
In a second step, the machine uses the detected patterns and compares them with other example data sets to check how precise the detected patterns and regularities are. This results in the so-called confidence level of the model. This helps to assess whether the data provided is sufficient or whether more or other data needs to be provided.
As soon as the confidence level is correspondingly high, in the third and final step new data sets that do not yet contain the desired information can be compared with the model and the desired information can be predicted.
The database determines the success of a machine learning project.
Since the examples used for pattern recognition are the most important element in an ML case, the database must be built up as early and as extensively as possible. Internet-of-Things scenarios are frequently used, which deliver masses of data from the edge, i.e. from a machine or a device, into a so-called Data Lake, i.e. a cost-effective way of storing terrabytes of data. Internet of Things is therefore not a mandatory component of an ML case, but it is often an enabler for an ML case, especially in predictive maintenance approaches.
SAP Cloud Platform Predictive Services offer a variety of options for predictive analytics
With the SAP Cloud Platform (SCP) it is possible to use Machine Learning without large upfront investments. The predictive services of SCP offer, for example, the possibility to detect outliers or to create forecasts, influence analyses and simulations.
Thus, with the predictive services of SCP, a multitude of different scenarios can be implemented within an ML project. For example, an online shop can analyze its previous returns and inform an employee about a possible return immediately after the purchase has been completed, so that the employee can contact the customer before the goods are shipped. A further scenario would be that the data from machines at a manufacturing company can be used to predict the next failure of the machine and thus trigger preventive maintenance measures. Companies with a large amount of master data can use SCP's predictive services to identify errors in price, customer or material master data. Machine learning can also be used to predict incoming cash flows and thus optimize liquidity planning.
Meinolf Schäfer, Senior Director Sales & MarketingDo you have any questions? I will be glad to help you.
+49 2241 8845-623