Good teacher for mining algorithms

The assignment of the bachelor thesis took place within the framework of the cooperation of the HdM with the company Uniserv from Pforzheim, a specialized provider of solutions for customer data management, which was agreed upon in the spring.

The author of the paper, Paul Titze, a student at the Department of Information and Communication in the Business Informatics and Digital Media program at the HdM, used various test scenarios in which data analyses were carried out using master data of different quality to check the correlation between high-quality master data and the results of the analysis via supervised machine learning.

Result: Particularly in supervised learning, in which the master data forms the basis for the learning of the algorithm, significantly better predictions could be achieved with a high-quality data basis prepared by master data management than with machine learning with an untreated data set.

"With his work, Mr. Titze has also demonstrated empirically for the first time that data quality and predictive analytics forecast quality are directly related"

"So companies using predictive analytics today as a basis for strategic business decisions should do everything they can to achieve the highest possible data quality in the data underlying the analysis."

Study course and prototype

The awarding of the bachelor's thesis is just one of the joint activities agreed upon as part of a cooperation between the Media University and Uniserv.

For example, the HdM has been offering the new master's program in Data Science and Business Analytics since the 2016/2017 winter semester. The continuing education program comprises five semesters and is part-time.

"The need for trained data specialists is enormous. International corporations in particular urgently need the know-how on how to generate knowledge from large volumes of data.

Practical scenarios exist in the automotive industry, for example, which is able to predict the wear and tear on a car based on a driver's driving style.

But data plays a major role in marketing in particular. The best-known example is the mail-order company Amazon, which provides customers with personalized offers based on their search and purchase history."

Uniserv supports the new training offer in terms of content and financially.

Golden Profile

For the training, the Media University uses the solution and process methodology Ground Truth, which was newly developed by Uniserv.

Ground Truth helps companies ultimately create each customer's Golden Profile in a multi-step process that aggregates their address data, buying behavior, interests and preferences, as well as their communications and interactions with the company into a central data set.

The two cooperation partners developed a prototype based on ground truth specifically for predictive analytics.

This prototype was intended to illustrate the importance of data quality as a critical success factor for the quality of forecasts.

Professor Lehmann explains the importance of ground truth as a basis for predictive analytics:

"Ground Truth is a didactically sound teacher of mining algorithms."