This week’s list of data news highlights covers December 5, 2020 – December 11, 2020 and includes articles about designing COVID-19 testing strategies with data analytics and better understanding how oceans absorb carbon with machine learning.
1. Using AI Matchmaking to Boost Birth Rates
The Japanese government is investing $19 million in AI dating services to bring single citizens together in order to boost the country’s low birth rate. The birth rate has been steadily declining in Japan, with the number of births dropping from 2 births per woman in 1960 to 1.42 in 2018. Japan has declared the situation a national crisis, prompting the government to issue subsidies to AI dating services that use data taken from standardized questionnaires, such as one’s political views, likes and dislikes, and goals, to connect Japanese citizens with their most compatible matches.
2. Designing COVID-19 Testing Strategies with Data Analytics
Massachusetts General Hospital has collaborated with researchers at the Massachusetts Institute of Technology to develop a tool that looks at how well companies are adhering to public health guidelines and generates customized COVID-19 mitigation strategies based on the risks they pose. The tool uses information about a company’s worksite, such as whether employees wear masks throughout the day or are social distancing, to inform them on the types of testing they should use. The tool describes the speed and accuracy of different testing methods, including onsite and lab-based testing, and also estimates the number of people the company should test each day.
3. Detecting Flooded Roads with AI
Researchers at Old Dominion University in Virginia are using AI to detect flooded roads from surveillance video footage. Researchers are using LiDAR, a type of laser technology that uses light to measure distances, to assess how deep floodwaters are and create a 3D map of the road. Using this map, researchers plan to build a predictive system that can simulate different flooding scenarios and alert drivers in real-time about problems on their driving route.
4. Predicting Which Breast Cancer Patients Need Surgery with Machine Learning
Researchers at the University of Michigan have developed a machine learning algorithm that predicts how the conditions of patients with early stage breast cancer are likely to develop, to inform whether they should undergo surgery to prevent the spread of cancerous cells. The team trained the algorithm to identify physical features of precancerous cells in breasts using medical images of patients’ breasts. The algorithm predicted cancer recurrences and nonrecurrences with 91 percent accuracy.
5. Identifying Brain Activity in People with Depression Using Machine Learning
Researchers from the Advanced Telecommunications Research Institute International in Japan have used machine learning to identify brain activity patterns in people with depression. The researchers trained the algorithm on the brain activity of 713 people, 149 of whom were diagnosed with depression. When tested, the algorithm identified people with depression with 70 percent accuracy.
6. Identifying Mild Cognitive Impairments with Machine Learning
Researchers at Queen’s University in Northern Ireland have used machine learning to identify Parkinson’s disease in patients with mild cognitive impairments, as well as in those without. Researchers trained the algorithm on electrical brain activity data from 100 patients, 50 of whom had Parkinson’s, to identify changes that occur when patients are engaged in tasks that require attention, language, and memory. When tested, the algorithm identified patients that have mild cognitive impairments with 80 percent accuracy.
7. Creating a Tool to Identify Traffic Congestion
Researchers at the U.S. Department of Energy’s Pacific Northwest National Laboratory have created a machine learning tool to identify real-time traffic patterns in cities. The researchers trained the tool on Uber driver datasets and traffic sensor data from the Los Angeles metropolitan area. The tool can model traffic congestion by the hour or minute.
8. Using Machine Learning to Better Understand How Oceans Absorb Carbon
Researchers and scientists at OcéanIA, a Chilean project developing tools to better understand the dynamics of oceans, are using machine learning to identify and document plankton, one of the largest carbon-capturing organisms on Earth. There are an estimated 70,000 unknown plankton species in the ocean today. The OcéanIA team are applying machine learning to a dataset of plankton images in order to identify these unknown species and better understand their unique features. With this method, the team hopes to better understand how the ocean absorbs carbon, produces oxygen, and supports biodiversity.
9. Transcribing Historical Documents Using AI
Amsterdam City Archives in the Netherlands has partnered with an EU-funded archival project called READ, to transcribe historical documents using AI. Amsterdam City Archives has a collection of documents approximately 50 kilometers long, which makes manually recording details from these documents an arduous process that requires decades of work and funding. Using READ, archivists can train the algorithm on existing transcriptions and have the algorithm compare the handwriting patterns it knows with that of other documents the user wants to transcribe.
10. Detecting Corruption with AI
Microsoft has partnered with the World Bank to detect corrupt behavior in financial transactions by applying AI to datasets released by international organizations and corporations. Using Microsoft’s AI tools, investigators will be able to better identify links in bidding patterns during contract negotiations and transfers of ownership. Additionally, investigators can better map out networks of locations, shell companies, and banking information that exhibit risky behavior.
Image credit: Jeffrey Hamilton
