AI-based Fraud Detection for Power Supply
Power-supply companies lose a lot of money mainly caused by non-technical losses, this is, those produced by customers who illegally intervene electrical meters (aka. Non-Registered Consumption or NRC). Business analysists struggle to detect those kind of behaviors in residential areas with low detection accuracy. In order to address this problem, ensemble learning using different types of machine learning techniques, was used to automatically predict NRC. Results using the AI model outperformed three or four times the human performance.
Improving Army's Integrated Support Systems using AI Technologies
Managing the sustainability systems of the military forces has several complex problems that affect the efficiency of its logistics processes. These include excessive time in the generation of purchase orders, poor reliability of the data recorded in internal systems, high costs due to machine maintenance downtime and/or defense assets, delays in procedures related to acquisitions and tenders, and increased transportation times, among many others. In order to address these difficulties, novel approaches and technologies such as AI and meta-heuristic optimization methods were proposed to solve some problems in logistic areas such as maintenance, transportation, procurement and planning.
Automated Answering for Contact Centers using NLP
Companies worldwide use call centers to interact with customers aiming to provide them with technical support, replies to questions/complaints, sales information, etc. Generally, decisions to satisfy users’ needs are made by human agents who are supported by state-of-the-art information technologies. However, this brings up a significant bottleneck involving productivity issues such as response times, quality of service, times of services, heterogeneity of responses, etc. In order to address these issues, this project aimed to develop an advanced intelligent model and computational prototype of a new approach to natural language text based call routing. AI and NLP techniques ae used to extract, identify and classify customers’ intentions and their ‘speech acts’ from textual information so to be routed to human agents.
Credit Card Fraud Classification
Recently, many people were victims of identity fraud, and credit card fraud continues to be one of the most common complaints by bank customers. Credit card fraud occurs when a third party makes a purchase using credit card information without the express permission of the account holder. Institutions such as banks and retail chains collect vast amounts of data in the course of doing business – historical customer data that can be used to detect fraud patterns to determine future probabilities. In order to address this problem, a combination of Neural Net based prediction methods were use to detect fraud on credit card transactions for a large retail chain, achieving almost 97% of accuracy. Although this predictive analytics does not reveal what type of fraud will happen, it will point to what might happen, so that actions can be taken to avoid and mitigate these events.
Intelligent Optimization to Reduce Waiting Times in Hospitals
One of the key issues with public health systems is the extremely long waiting times in state hospitals both for outpatient and inpatient health care, hence novel procedures, methodologies and approaches are required to reduce critical times. In order to provide automated support for decision makers in hospitals and to reduce human intervention in outpatient care, intelligent optimization technology and machine learning techniques were proposed so as to make hospital management and resource allocation more efficient, reducing waiting times on primary health care.
Tourism Data Observatory
There is a lot of uncertainty in the tourism industry that directly affects business decisions, public policies, and competitiveness. In order to contribute to improving the competitiveness of the tourism industry, a tourism intelligence observatory was designed and implemented based on large amounts of data analytics technology, which allows characterizing the consumption behavior of the international market and promoting prospective planning. Analytics dashboards were built so as to provide insights on consumer profiles, travel satisfaction, purchasing behavior, projection of visits, markets of interest, etc.
Intelligent Media Exploration & Analysis for Intelligence Tasks
There are multiple sources of information that can account for the behavior and preferences of users about certain events of national interest, both from formal media (e.g., documents, institutional information records, geo-location information) and informal (e.g., social networks, electronic forums, newspapers). All together becomes a very relevant strategic information that can be used for decision making, exploring and predicting future behavior or detecting certain patterns in events. To address these issues, various AI prototypes were developed to support intelligence analysis tasks. This made it possible to answer several questions related to user preferences in social networks, major influence hubs, tracking conversation topics, etc.