Table of Contents
Switzerland, Germany and Austria are a little hesitant about the use of new technologies – it has been thought so far, and this cliché persists. In practice, however, this seems to be different in German companies: more than two-thirds of all German companies have already used artificial intelligence (AI) and machine learning, or at least are in the process of implementing corresponding projects. More than 20 percent of all companies even have multiple machine learning applications in operation. This is the result of a study carried out by the market research institute IDG Research in 2019. But what is machine learning and what can it be used for in a value chain? You can find out more in this post.
How Machine Learning (ML) Works
Machine learning is a subset of artificial intelligence. Networked IT systems automatically learn from experiences, such as recurring data patterns. Algorithms are developed for this purpose for the respective task. In other words, machine learning is used to develop programs that in turn access data to detect patterns. A common field of application is the evaluation and analysis of large amounts of data in the field of data mining.
For machine learning (ML) to work at all and for the programs used to be able to make decisions, the associated algorithms must be trained by humans. For this purpose, example and training data are used, in which the algorithm recognizes relationships and patterns, learns from them and draws conclusions. This process is also known as model training.
Once usable training data is available, machine learning is able to:
- Probability calculations for specific events
- Recognition of correlations in present sequences
- Compressing dimensions without significant information loss
- Make forecasts based on the analyzed data
- Business process optimization in the business environment
Leverage machine learning in the enterprise
According to the above-mentioned IDG study, the greatest benefit of ML benefits is to IT departments and customer service. This is due, among other things, to the fact that security-related applications in particular use AI-based learning capabilities to detect spam, unusual behavior of network components, and cyberattacks. The same applies to applications that are used for the authentication of IT users – in particular, the evaluation of biometric data (face recognition, voice analysis, etc.)
AI applications are also playing an increasing role in industrial production environments. Network-based manufacturing systems, for example, use programs for condition monitoring and predictive maintenance. At the Hannover-Messe 2019, corresponding solutions were presented and impressively proved that machine learning has what it takes to open up new business areas – Smart Factory and Industry 4.0 are the common keywords here.
How do you enter the ML world as a company?
What should a company take to take the first steps in artificial intelligence and machine learning? Here, it only helps to approach corresponding projects in a practical way. Simply do it and gain experience is the motto here. However, in this context, the human factor should not be forgotten. The employees have to go along with it, if necessary fears and prejudices have to be reduced. Experts agree that machine learning only has a chance in the medium term if it is perceived as supportive. Technology as a threat is not a good guide here.
Another weakness that often makes it difficult to get started with AI and ML is simply the lack of IT infrastructure. This means not only high-performance hardware, but also cleanly managed data sets and, last but not least, lived data protection and data security. Many companies are actually still at the very beginning here.
The machine learning process in detail
In practice, the course of a machine learning process is as follows:
- Definition of the problem,its solution and the associated exchange of knowledge. The goal to be achieved using ML must be clearly defined in advance.
- The most complex step is data acquisition, feature extraction, and transformation. Here, it is essential to use high-quality data sets. An ML feature store can create high efficiency here.
- In the learning phase, the actual machine learning takes place, the ML algorithm starts its work with the training.
- Evaluation and analysis of the results. Here, model interpretation is an important step for the overall process. This is where the acceptance of machine learning in the specialist departments is to be achieved. Employees should recognize the benefits and understand what happens in detail in the algorithm.
- Only productive use ensures that ML creates real added value. After all, it is a useless investment to develop machine learning models and then not to use them. However, the complexity of the technical requirements sometimes ensures that the production of the system is not always easy.
What is Deep Learning?
In the context of machine learning, the term “deep learning” is used again and again. This is an ML subset, basically deep learning is always machine learning from a technical point of view. The differences lie in the capabilities: Deep learning is able to process unstructured data – sounds, texts, images, etc. – using artificial neural networks and convert it into digital data. This preparation data is then processed in the ML for further learning or pattern recognition. Traditional machine learning processes are unable to process unstructured data with their decision tree procedures. For example, an algorithm can not be trained with pure image data. Here, in turn, humans would have to perform time-consuming feature engineering as a tedious intermediate step. Deep learning takes this task away from him.
Practical applications for ML
In our digitized world, machine learning finds a place almost everywhere. The following application examples are intended to illustrate this:
- Each search engine uses complex algorithms to generate the desired information and learns from the search queries and results each time.
- In e-commerce, many projects are supported by recommender systems for personalizing the purchasing experience.
- In medicine, scientists and physicians can use analytical models to identify disease-specific risk factors.
- In CRM, ML helps improve marketing campaigns, predict cancellations, or determine a customer’s value.
- Every user of Facebook, YouTube or Instagram knows that the content intended for him is individually tailored to his or her needs by analytical models. ML identifies personalized interests in the social networks and evaluates behaviorpatterns.
Conclusion: There is still a lot to do
In summary, it can be said that, according to the IDG study from 2019, German companies are already on the right track in terms of acceptance of machine learning and AI. Even companies of the Size of SMEs deal with this issue and do not leave it to the big players. But even a not insignificant number of companies are still sceptical about the matter or have not yet understood the strategic importance of ML and AI. In many cases, these technologies are deficiently or not included in the business models. There is a clear need to catch up here. So there is still much to be done, otherwise German companies are in danger of being left behind in the global market.