HONG KONG, June 16, 2026 /PRNewswire/ — MetaLight Inc. (“MetaLight” or the “Company”; Stock Code: 02605.HK), a public transit information service provider powered by time series data analytics and AI technology, today announced that a peer-reviewed research paper co-authored with Peking University and other institutions has been accepted to the Applied Data Science (ADS) Track of KDD 2026. Titled “A Data-driven Route Segmentation Framework for Time-of-Arrival Estimation Service,” the paper’s authors are affiliated with Peking University, MetaLight HK Limited (a member of the Group) and other institutions, and its first author is from Peking University; the Company’s Chairman and CEO, Dr. Sun Xi, is a co-author. It addresses route segmentation — a previously underexplored step in bus arrival-time estimation (ETA); the Company’s team, drawing on its Chelaile platform, provided the real-world data and the online deployment environment for the work.
KDD (the ACM SIGKDD Conference on Knowledge Discovery and Data Mining), organized by ACM SIGKDD, is rated a Class A recommended conference by the China Computer Federation (CCF) and will be held in Jeju, South Korea, from August 9 to 13, 2026. Accepted in the track’s second review cycle (Cycle Two), the paper will be included in the proceedings and presented at the conference by its authors.
In bus ETA, a route is typically divided into segments using predefined markers such as intersections, with traffic data aggregated accordingly. Coarse segmentation can blur fine-grained differences in traffic conditions along the same route and constrain prediction accuracy. The research instead draws on large-scale, fine-grained vehicle trajectory data to segment routes so that each segment corresponds to a relatively stable traffic regime, characterizing real traffic variation along a route more precisely.
In offline evaluations covering over a thousand routes and more than a million trajectories, the method outperformed commonly used route-segmentation strategies. In a one-week online evaluation on the Chelaile service across two cities, it reduced storage usage by approximately 90% and CPU usage by approximately 25% at comparable prediction accuracy.
Prior ETA research has focused largely on the prediction model itself, with less attention to how a route is segmented and how its data is organized; this work indicates that this step is just as important to prediction accuracy and operational efficiency. In the Company’s bus ETA service, the approach has materially lowered storage and compute requirements at comparable accuracy, making the large-scale online service lighter to run.
Beyond the work accepted to KDD 2026, MetaLight has in recent years also co-authored papers in temporal point processes (with Renmin University of China, Shanghai Jiao Tong University and Fudan University, accepted to AAAI 2026) and model merging (with Shanghai Jiao Tong University and the Shanghai Artificial Intelligence Laboratory, accepted to IJCNN 2025).
A representative of MetaLight said, “This work benefited from the collaboration of our team with Peking University and other partners. We care more about whether technology solves concrete problems in real-world settings, and we also hope to foster an internal culture of active learning and continuous evolution; we will keep building our foundational technology steadily, through long-term investment and collaboration.”
About MetaLight
MetaLight Inc. (Stock Code: 02605.HK) is a public transit information service provider powered by time series data analytics and AI technology. With a focus on time series data foundation models, the Company has built a technology stack with its core comprising an AI Model Building Platform and AI model libraries for three industry verticals: public bus, renewable energy and industrial internet, integrating capabilities in data access, pre-processing, labeling, AI model training and foundation model adaptation. Based on this technology stack, the Company operates the Chelaile real-time public transit information platform, providing commuters with services including real-time bus arrival predictions, vehicle location tracking and travel route planning, while also offering public transit analytics platforms and data technology services to transport operators. According to CIC data as of December 31, 2024, Chelaile is the largest real-time public transit information platform in China by city coverage. As of December 31, 2025, it covered 488 cities and towns nationwide with approximately 334 million cumulative users, committed to making public transit more convenient and efficient. For more information, please visit www.metalight.ai.
Forward-Looking Statements
In addition to statements of historical fact, this press release contains forward-looking statements relating to the Company’s business outlook, estimates of financial performance, forecast business plans, development strategies and projections of anticipated trends in our industry. Forward-looking statements can generally be identified by the use of forward-looking terminology such as “may,” “might,” “can,” “could,” “will,” “would,” “expect,” “believe,” “continue,” “estimate,” “anticipate,” “forecast,” “intend,” “plan,” “seek” or “timetable.” Such forward-looking statements are based on the information available to the Company and the outlook as at the time of publication of this press release. Such forward-looking statements are based on certain projections, assumptions and premises, some of which involve subjective factors or factors beyond our control. Such forward-looking statements may prove to be inaccurate and may not materialize in the future. Such forward-looking statements involve a number of risks and uncertainties. In view of the risks and uncertainties, forward-looking statements contained in this press release should not be construed as representations by the Board or the Company that such plans and objectives will be achieved, and investors should not place undue reliance on such statements. Except as required by law, neither the Company, its Board, employees nor agents assume any obligation to publicly release any revisions, corrections or updates to the forward-looking statements contained in this press release to reflect events or circumstances occurring after the date of this press release or unanticipated events, nor do they assume any liability for any losses arising from the failure to realize or inaccuracy of any forward-looking statements.
Investor and Media Inquiries
MetaLight Inc.
Investor Relations
Email: ir@metalight.ai
