The click-through rate (CTR) is a ratio that represents how often people click on a specific digital resource, such as a blog entry, ad or hotel promotion. Accurately estimating CTR is critical for determining the price we are willing to pay for ads or deciding on the user segment we want to target. We therefore designed a real-time CTR prediction algorithm for an online travel company based on the analysis of millions of high-dimensional data items representing its customers’ behaviour.
It is said that data is the oil of the 21st century, but in order to unlock its potential, it is important to analyse it carefully and extract knowledge from it. In digital environments, data science can be used to identify trends, public opinion and communities. In media environments, it is possible to go a step further and profile audiences and users with a view to offering them content that responds to their interests. In banking environments, data can reveal a great deal about the habits of customers and make it possible to offer them services tailored to their needs. In healthcare environments, data can be used to improve patient care and make patients the focal point of the healthcare system. In environments related to organisation management, careful examination of internal communications allows the strengths and weaknesses of an organisation to be identified.
- 1. DIGITAL ENVIRONMENT (mobile, internet and social networks)
- Monitoring and analysis of social networks.
- Positioning and improvement of visibility.
- Building customer/user loyalty.
- Online reputation.
- Identification of trends.
- Recognition of communities.
- Segmentation of users.
- Mobility and human behaviour studies in controlled environments.
- CTR forecasting.
- Behaviour-based recommender systems.
- 2. MEDIA (applied to content management)
- Content-based collaborative recommender systems.
- Indexing and generation of metadata.
- Segmentation of users and content.
Image search engines receive huge volumes of images every day, but very few of these have any impact on the internet. Of course, the aesthetics and technique of images are key to ensuring they have an impact on the internet and are not overlooked.
- 3. BANKING, INSURANCE AND FINANCIAL MARKETS
- Solutions for getting to know the customer.
- Descriptive and predictive alerts regarding financial incidents.
- Fraud detection.
- Risk measurement.
- Optimisation of decisions and strategies.
- Analysis and segmentation of customers’ financial behaviour.
- 4. HEALTHCARE
- Techniques for making clinical, medical and management decisions.
- Monitoring of patients.
- Smart health warnings.
- Medical risk assessment.
- Monitoring of patients’ recovery.
- Segmentation of chronically ill patients.
- Analysis of medical imaging and data.
Development of a system for diagnosing intestinal motility disorders based on images taken by a wireless endoscope camera that describes a range of visual patterns, both static and dynamic, from which an objective diagnosis can be made.
- 5. HUMAN RESOURCES AND ORGANISATION MANAGEMENT
- Analysis of organisations.
- Efficiency of organisational processes.
- Worker relations.
- Identification of emerging groups.
- Measurement of cohesion among work teams.
- Analysis of organisational and social networks.
This project consisted of the development of a tool for learning about the different relationships between the members of an organisation and their areas of work. The group-level analysis provides a real-time diagnosis and makes it possible to take actions to improve and streamline processes.
Other solution cases
We have won a new competition with the Cartographic and Geological Institute of Catalonia, to apply deep learning techniques in the detection of changes in the network of infrastructures and constructions in the territory with the aim of automating the process of elaboration of the Topographic map of Catalonia.
Joint development with Vall Hebron Hospital of a scalable medical diagnosis solution based on artificial intelligence tools for CorporateHealth International ApS.
DataScience@UB designs and prototype a machine that includes an intelligent vision system based on semantic segmentation through convolutional neural networks to automate the classification of leather.