This week’s list of data news highlights covers April 3, 2021 – April 9, 2021 and includes articles about using open data to help countries address extreme climate events and predicting stock market crashes.
1. Using Data from Pedestrians and Cyclists to Improve Urban Planning
Strava, an app that tracks exercise activity using GPS data, has created anonymized datasets about the activity of pedestrians and cyclists, such as the paths they frequently use, to improve urban city planning. So far, the city of Frankfurt, Germany has used data from the company to determine how pedestrians and cyclists travel in the city center and decided to drop a section of a curb on a busy footpath to increase access for cyclists.
2. Forecasting Surges of Coronavirus
The National Minority Quality Forum, a non-profit and non-partisan education organization, has launched the COVID-19 Index, a predictive analytics tool that uses historical data about coronavirus infections rates by zip code and regional areas, to forecast future surges in infection rates in underserved communities. The index also includes a disease-specific map, which allows local officials to visualize localized viral trends before they reach a critical stage so that resources are appropriately prioritized and allocated for vulnerable populations.
3. Creating a Dataset for Food Producers
Researchers from Saarland University in Germany have developed Evarest, a data platform used to improve food production. Evarest collates anonymized operational data from food producers including production quantities and volumes, weather conditions that may impact crop production, and plant or equipment production from other companies. The tool uses AI to process and link different datasets, allowing producers to better optimize their production processes, such as by controlling production volumes to prevent overproduction.
4. Evaluating the Performance of AI Systems for Bias
Facebook has created Casual Conversations, an open-source dataset comprising 45,000 videos of 3,000 people having non-scripted conversations. The dataset is evenly distributed across different genders, age groups, and six skin types. It also includes labels of the levels of light in each video to help researchers evaluate how well AI systems treat people with different skin tones in lower-light conditions and prevent bias.
5. Predicting Stock Market Crashes
Scientists from the École Polytechnique Fédérale de Lausanne, a research institute in Switzerland, are using topological data analysis, a technique that extracts information from large volumes of data, to predict when the stock market will crash. By analyzing daily price data from the S&P 500, an index commonly used to benchmark the state of the financial market, the team created a price-based graph that showed numerous peaks exceeding the warning levels for the market’s instability in the months leading up to the crashes in 2000 and 2008.
6. Identifying Subtypes of Multiple Sclerosis from Brain Scans
Scientists at the University of College London have used AI to identify three subtypes of multiple sclerosis (MS) from brain MRI scans. The team trained an AI tool to identify patterns of brain matter from 6,300 brain scans of MS patients. When they applied the tool to 3,000 new MS patient scans, they discovered three previously unknown subtypes. Their findings confirmed that patients with an MS subtype defined as lesion-led had the highest risk of disability progression and relapse rate.
7. Using Open Data to Help Countries Address Extreme Climate Events
Countries are using open data tools to address the impact of extreme climate events, such as flooding and heatwaves. In the United Kingdom, the city of Manchester has developed a tool that combines data from government and academic sources to visualize how heatwaves, flooding, and other extreme climate events will affect 1,000 impoverished communities. In Malabon City in the Philippines, local officials used data from the Philippine Atmospheric, Geophysical and Astronomical Services Administration to identify the most at-risk neighborhoods and create flood risk maps to encourage community participation in developing evacuation plans.
8. Predicting Changes in Sea Levels
Researchers at the University of Valencia in Spain have developed a machine learning approach that uses estimates of sea temperatures to model the variability of sea levels in the Pacific, Indian, and Atlantic oceans across different time ranges. The models could better explain how natural processes in large open oceans, such as temperature changes, influence short-term variations in coastal sea levels, which in turn influence other processes such as high tides or storms.
9. Analyzing Baseball Pitches with a Supercomputer
Researchers at the Tokyo Institute of Technology have used supercomputer simulations to discover why forkball baseball pitches, which are baseball pitches where the pitcher snaps their wrist down, moves the ball so far downward that it disappears from the batter’s point of view. The team simulated airflows associated with forkballs on a supercomputer and discovered that unlike straight-ball pitches where air flies upwards, the airflow from a forkball pitch exerts a downward force on the ball during one-third of its rotation. When the team compared straight-ball pitches to forkball pitches delivered with the same speed and number of rotations, the forkball reached a catcher up to 19 centimeters lower.
Google is using AI and augmented reality (AR) to include new features on Google Maps. Users can now virtually navigate indoor facilities, such as airports, shopping malls, and train stations, to locate elevators, escalators, ATMs, and waiting platforms. Google has also partnered with the U.S. Energy Department to use AI to create green routes, which limit a user’s carbon footprint for the same initial travel time.
Image credit: Jamie Street