Expanding use cases for Prophet
Team members
Jawad Jafar
Mandy Alevra
Leeah Michael
Remy Wu
Faculty mentor
Markus Pelger
Sponsor organization
Prophet is a startup that has developed a general-purpose decision-making algorithm.
Project description
Our project set out to assess the potential of Prophet's machine learning algorithm beyond its proven success in the digital marketing realm, specifically targeting the urban planning sector. The goal was to explore various subcategories within urban planning to identify new markets where Prophet’s technology could be applicable. By analyzing the demands and conditions of each sector, we aimed to establish a strategic framework for Prophet to expand its use case portfolio and prepare for public API licensing.
The central challenge was to find suitable sectors with relevant datasets. We also evaluated the practicality of Prophet's reliance on simulated data when applied to real-world scenarios, with a particular focus on traffic control systems. This involved gauging the algorithm's performance in simulations against the unpredictable dynamics of actual urban environments. Our goal was to validate the adaptability of Prophet's technology to complex, real-world MPC problems.
Techniques and methods used
To assess Prophet's applicability, we employed a multi-faceted approach that began with data collection and analysis. We scoured public databases for datasets that could potentially align with Prophet's machine learning capabilities, focusing on the frequency of data points, variety of variables, and the presence of actionable switches.
We utilized the Simulation of Urban Mobility (SUMO) platform to analyze traffic flow, which provided a controlled environment for testing Prophet's algorithm. This simulation approach was invaluable in allowing us to gauge the algorithm's decision-making and predictive capabilities without real-world risks.
However, the reliance on simulated environments presented challenges. While SUMO provided a robust framework for initial testing, the transition to a real-world application was not straightforward. The discrepancies between simulated and actual data highlighted the need for more nuanced data that could reflect the complexities of urban systems. Consequently, while simulations were crucial in early testing, they were insufficient for full-scale application, underscoring the need for real-world pilots to validate Prophet's technology further.
Throughout our research, we found that while Prophet's technology showed promise in theory, the practical application required overcoming significant barriers, such as data accessibility and integration into existing urban systems. Identifying and accessing comprehensive real-time data proved to be a significant hurdle. Going forward, the emphasis would need to shift from simulated to actual environments to fine-tune the algorithm's effectiveness in real-world scenarios.
Solutions and deliverables
Our primary deliverable was a comprehensive analysis report that pinpointed traffic flow optimization as a viable sector for Prophet's technology. We provided Prophet with a detailed case study, illustrating how their algorithm could be adapted to manage and optimize urban traffic flows. The simulated datasets from SUMO, transformed to be compatible with Prophet’s API, offered a preliminary proof of concept for how their tool could predict and alleviate traffic congestion.
This case study, coupled with a proposed framework for sector analysis, is intended to guide Prophet as they consider expanding into new markets. The recommendations and criteria outlined in the report serve as a strategic tool for Prophet to identify economically viable and technologically suitable markets with low barriers to entry and potential for impactful application of their machine learning algorithm.
Prophet plans to leverage our findings and recommendations to refine their expansion strategy and develop a broader use case catalog for their API. By following our outlined approach, they can systematically assess new opportunities, prioritize sectors for entry, and ultimately advance towards public licensing of their API.