Researchers use information extracted from tweets to provide unmatched accuracy in predicting morning traffic patterns.
Morning commute is one of the busiest hours. However, it has also proven to be the most difficult time to predict traffic patterns. This is because most methods of traffic forecasting rely on having a consistent flow of traffic data from the time to the forecast period.
But the vast majority of people spend time in front of them Commuting Sleeping at home or doing your morning routine leaves a big gap in your forecast traffic data.
The researcher’s method solves this problem by retrieving data from tweets sent between the evening of the previous day and the early morning of the next day. They first used Twitter’s application programming interface (API) to identify tweets within a particular area (in this case, the city of Pittsburgh) by geotagging to indicate the source. Then use another application called Twint, a web scraper, to get other posts from the user with geotagged tweets and create a better image of the time the user was active and the general area. Did. All data has been anonymized and personally identifiable information has been removed prior to publication.
“Tweet claims to capture three types of useful information to explain traffic the next morning, including people’s sleep wakefulness, community events, and (planned) traffic accidents,” the author said. Written by Sean Qian, Associate Professor of Civil and Environmental Engineering, and Weeiran Yao, a PhD student in Qian.
By further expanding this dataset, researchers were able to extract additional information. The team used linguistic analysis to identify search terms that could indicate a traffic accident. This includes not only accidents, but also large-scale events such as planned closures, concerts, sports games and holiday celebrations.
Simple personal tweets like “Great excitement in the Pirates game!” Or “This fender vendor ahead will delay me”, especially tagged with geotags or other tweets from that user. If notified by, you can actually provide important information. In addition, data was extracted from official accounts such as the press and local governments, who often tweet their direct reports regarding accidents and planned closures.
The combination of these methods provides a large dataset of information that indicates the likely geographic distribution and sleep / wake time. CommuterBoth planned and accidental road accidents that can affect commuting. This filled the information gap created by the drop in traffic at night.
With this information, Qian and Yao can provide traffic forecasts for Pittsburgh’s morning commute with unprecedented accuracy, creating a comprehensive framework for forecasting urban morning traffic conditions. Did. area..
This information also allows them to begin observing and predicting on a larger daily scale. It turns out that Pittsburgh morning traffic is generally crowded on Tuesdays, Wednesdays and Thursdays. This will allow transportation to better manage its morning commute. These types of observations were previously impossible because they cannot accurately predict morning conditions, but they can help make larger decisions in travel demand management, signal timing control, and personal destination routing.
“This study leverages machine learning and big data to understand human behavior while protecting personal privacy,” says Qian.
“It’s very exciting to see this method lead to a better forecast of morning commuting traffic by 5am. I believe this can be quickly implemented in many traffic management centers.”
Those results will be displayed in Traffic research..
https://www.nextgov.com/analytics-data/2021/05/tweets-fuel-super-accurate-morning-traffic-prediction/173984/ Tweets facilitate super-accurate morning traffic forecasts