The role of AI and advanced algorithms in urban development is rapidly expanding, bringing transformative changes to how communities are planned and managed. A new wave of AI-driven tools, particularly those based on transformer models, is revolutionizing time series forecasting in urban planning. These models are proving crucial for predictive accuracy in managing growth, especially in dynamic environments like master-planned communities.
What if we could integrate Time-Varying Markov Models (TVMM) with AI to enhance forecasting precision? A recent paper exploring dynamics of growth of master-planned communities highlights the importance of incorporating dynamic, data-driven approaches to forecasting housing growth in master-planned communities, laying the groundwork for advanced AI-driven models that can further enhance our understanding of housing development patterns.
As these communities evolve, AI-driven predictions will become increasingly vital for sustainable growth, efficient resource allocation, and enhanced quality of life.
Among the most popular time series transformers in time series data (that could be extended to urban planning) are foundation models like Chronos, TimesFM, Moirai, and TimeGPT. Each model offers unique strengths that cater to different forecasting needs:
Chronos: Developed by Amazon, this open-source model treats time series as specialized languages with their own patterns. Despite its simplistic approach, Chronos has shown impressive results across various forecasting scenarios, making it a reliable tool for general-purpose forecasting.
TimesFM: Created by Google Research, TimesFM is trained on over 100 billion real-world time series points. This model allows fine-grained control over seasonal patterns and has proven to be a powerful and flexible forecasting tool, especially in complex urban settings.
Moirai: From Salesforce AI Research, Moirai is designed to handle both missing values and external variables, making it a versatile choice for urban planning. Its ability to adjust to different seasonal patterns makes it an invaluable tool for forecasting in diverse environments.
TimeGPT: A proprietary production-ready model, TimeGPT excels in ease of use and supports external variables. It’s particularly effective for organizations needing quick, reliable forecasts with minimal setup. Its performance across a wide range of time series data underscores its value in fast-paced, real-time applications.
As we look to the future, these AI-driven models will play a pivotal role in shaping the growth of our communities. With tools like TVMM and advanced transformers at our disposal, urban planners can make more informed decisions, ensuring that the communities of tomorrow are both sustainable and resilient.
REFERENCES
Christopher K. Allsup, Irene S. Gabashvili. Modeling the Dynamics of Growth in Master-Planned Communities August, 2024 arXiv:2408.14214 [econ.EM]
Abdul Fatir Ansari, Lorenzo Stella, Caner Turkmen, Xiyuan Zhang, Pedro Mercado, Huibin Shen, Oleksandr Shchur, Syama Sundar Rangapuram, Sebastian Pineda Arango, Shubham Kapoor, Jasper Zschiegner, Danielle C. Maddix, Hao Wang, Michael W. Mahoney, Kari Torkkola, Andrew Gordon Wilson, Michael Bohlke-Schneider, Yuyang Wang Chronos: Learning the Language of Time Series arXiv:2403.07815 [cs.LG] https://doi.org/10.48550/arXiv.2403.07815 [Submitted on 12 Mar 2024 (v1), last revised 2 May 2024] Code and model checkpoints available at https://github.com/amazon-science/chronos-forecasting
Abdul Fatir Ansari, Lorenzo Stella Adapting language model architectures for time series forecasting March 18, 2024. Amazon Science Blog
Abhimanyu Das, Weihao Kong, Andrew Leach, Mike Lawrence, Alex Martin, Rajat Sen, Yang Yang, Skander Hannachi, Ivan Kuznetsov and Yichen Zhou. https://research.google/blog/a-decoder-only-foundation-model-for-time-series-forecasting/
Gerald Woo, Chenghao Liu, Akshat Kumar, Caiming Xiong, Silvio Savarese, Doyen Sahoo. Unified Training of Universal Time Series Forecasting Transformers arXiv:2402.02592 https://doi.org/10.48550/arXiv.2402.02592
Azul Garza, Cristian Challu, Max Mergenthaler-Canseco TimeGPT-1 arXiv:2310.03589 https://doi.org/10.48550/arXiv.2310.03589
No comments:
Post a Comment