Machine Learning in FinTech | Document scanning
One of the
leaders in the European financial sector was looking for a solution that would
support and improve its document flow system. Until now, each document sent to
the bank was scanned and sent to the right employee by the administration
department. However, this process took up too much time.
The goal of
the project was to create a system that could perform document analysis for
employees, then classify the document on the basis of the data read and assign
it to a specific department and person in the company.
We designed a ML solution based on an algorithm for image classification and transfer learning. The algorithm for classifying images made the recognition of text from document scans possible, which was followed by analysis for characteristic parts of the text. The transfer learning algorithm ensured the proper learning process of the entire solution based on the knowledge acquired during work.
Machine Learning in banking sector | Intelligent search engine for the chatbot
bank used a chatbot which featured a search mechanism based on keywords that was
used to handle some of the queries. In order to get an answer to a query, the
client had to enter specific phrases in the search engine. This meant that the
knowledge base required significant involvement from the consultant side. To
reduce the volume of this type of query, the bank decided to create a new
assumption of the project was to eliminate the simple search and replace it
with more advanced mechanisms based on semantic searching.
designed solution is based primarily on NLP (Natural Language Processing)
algorithms. The new chatbot expands its knowledge base using deep learning
algorithms. Thanks to this, based on the analysis of the answers given by
consultants, the system generates answers to previously unknown questions.
Machine Learning in Industry 4.0 | Analysis of the wear and tear of industrial elements using image analysis
industrial company that produces mechanical components needed a way to optimize
the exchange of cutting tools on CNC machines. In the event that the head was
replaced too late, the material being treated was damaged. In turn, if replaced
too early, the replacement generated additional costs associated with the
purchase of new tools.
The plan was to determine the best time for the service by means of a Machine Learning solution. In order to do so, a device was created to take a photo of the head after each work cycle. The algorithm used image analysis to determine the rate of wear and tear of the head in a given cycle. On this basis, it forecasted whether the given head was suitable for operation in the next cycle or whether it had to be replaced.
The use of a solution based on Machine Learning technology translated directly into a reduction of costs and the maximum use of the purchased tools.