Machine Learning & Data Science

↑

You hear about machine learning again and again, but rarely can you imagine how to make use of it and how you and your company can benefit. Whereas until a few years ago the topic was dealt with exclusively in universities or tech companies, today the use of artificial intelligence (AI) is becoming increasingly commonplace thanks to technological progress. Functions such as the speech recognition of digital assistants, spam filtering of e-mails, the targeting of personalized advertising on the Internet or medical diagnostic procedures now use machine learning as a matter of course.

The fashion and lifestyle industry uses artificial scarcity, also known as drop, to boost sales and provide customers with an exclusive brand experience. Resellers often exploit this to resell the sneakers – to continue with the example – for several times their original value.

Already sold out? How bots derail limited sales

Felix Martin Siepmann

26.11.2020

3 minutes

How insurers can use their large data quantities efficiently

Big data and the optimization of dunning processes

Fabian Küchlin

14.06.2018

4 minutes

How insurers can use their large data quantities efficiently

Machine learning refers to the general generation of knowledge from collected experience. It thus belongs to the data sciences.

Artificial intelligence

With the help of existing test data and rules for data analysis specified by humans, a system can build up a statistical model. In this way, it "learns" to recognize certain patterns and regularities in the data, and a kind of artificial intelligence (AI) emerges. In the best case, the system is also able to correctly assess new, unknown data. This should enable the system to analyze data or solve problems independently.

For example, a learned system can independently identify and extract relevant data, make predictions based on the database and thus calculate the probability of an event.

Partially supervised learning

only a part of the answers are known

Reinforcement learning

Rewarding desired behaviors

Active Learning

Possibility that the system asks questions

Independent learning

Algorithm acts as teacher and student

Machine learning can be roughly divided into two variants: supervised learning and unsupervised learning.

Supervised learning

In supervised learning, the system is taught a basic knowledge. This is done with training data that contains both an input parameter and the correct result. From this, the system can create example models for orientation that lead to the appropriate result. These algorithms can then be used to correctly categorize unknown data.

Possible sub-categories of supervised learning include:

In unsupervised learning, the system does not know what to recognize in the data. It forms independent model groups and divides the recognized patterns into clusters, but does not know which label matches them.
Unsupervised learning distinguishes between

Batch learning - all training data is given at once
Continuous learning - the model is built step-by-step

The uses of machine learning in enterprises are as varied as the different tasks performed by employees and the data that accumulates in modern enterprises.

Generally speaking:

The larger the database, the more reliable the results.

With the support of machine learning, humans can work more efficiently, because time-consuming, unloved work steps can simply be left to the computer. In most companies, for example, there are many man-hours spent on invoice verification, payment processing and back-office activities that can be automated using Robotic Process Automation (RPA).
Especially when it comes to processing large amounts of data, a learning program often even outperforms humans by proceeding faster and with fewer errors. Data science approaches make use of large database structures that are made accessible and interlinked with the help of programming languages such as Python.

Combating cybercrime and installing reliable chatbots

Machine learning is also used in the finance sector in cases of suspected credit card fraud. Especially in times of growing cybercrime, real-time fraud prevention is needed to detect suspicious activities in time while ensuring customer satisfaction through fast processing.

Customer Communication

Even the dunning process can involve the use of machine learning to find out which customers are best reached, when, in what tone and via which channel. Deriving the best measures can help to maintain the customer relationship despite payment arrears and improve the company's liquidity at the same time.

Artificial intelligence also finds application in the form of chatbots, as these can independently answer a large proportion of customer questions. Instead of having to wait longer for a free employee, prospective customers receive immediate feedback. This promotes satisfaction and relieves the burden on employees. A real win-win situation for both sides.

As the various application examples show, support from artificial intelligence brings many advantages for a company:

Shorter processing times reduce costs and increase customer satisfaction.
Employees can be deployed more efficiently, as repetitive tasks are handled by the computer.
Artificial intelligence means that a system is constantly "learning" and continuously minimizes the susceptibility to errors.
Systems can be individually and flexibly adapted to the structures of the company.

Thus, the targeted use of artificial intelligence can become a real competitive advantage. Learn more about Arvato Financial Solutions' experience with artificial intelligence and how we apply the technology.