City governments have long had to worry about keeping day-to-day life running for their constituents. Everything from fixing potholes to preventing crime have long been the success metrics for locally elected officials. In my opinion, this is why city governments are now jumping at machine learning tools that make all of these tasks easier (and cheaper).<\/p>\n
The Boston City Government, led by Marty Walsh, has invested heavily in machine learning technology to improve city processes. Frost & Sullivan ranks Boston in the Top 30 cities across the globe for using machine learning, specifically based on metrics like citizen-centric data analysis, private sector partnerships, an data aggregation across multiple sources.[i]<\/a> I would argue that Boston\u2019s investments in machine learning have been driven by two main factors: (1) machine learning is a low risk initiative given the city\u2019s access to a wide range of data, and (2) if done successfully, there are proven results that align with any politician\u2019s key success metrics of improving the quality of life for residents and saving tax payer dollars.<\/p>\n In general, cities have access to a wide range of data. In 2017, 2.3 billion connected devices and sensors were used across global cities (a 42% increase from 2016) as a means to capture data for machine learning.[ii]<\/a> Furthermore, traditional data collection at the city level (i.e. building details, restaurant registrations, etc.) can be used for machine learning algorithms to predict housing fires, restaurant health code violations and more.[iii]<\/a> As non-profit entities, I would argue that cities are also more effectively able to collect data from constituents. For example, Boston launched a Safest Driver competition with a free mobile app that analyzed driving behaviors, which led to valuable insights on features that predict unsafe driving in cities.[iv]<\/a> Passive data like this has been proven to be helpful in predicting accidents on city roads.[v]<\/a><\/p>\n These machine learning efforts help cities direct limited resources more efficiently and have tangible results. McKinsey Global Institute estimates that machine learning can help cities reduce crime incidents by 30-40%, help first responders arrive anywhere from 2-17 minutes earlier, shorten commute times by 15-20%, and potentially reduce the cost of living by 1-3%.[vi]<\/a><\/p>\n Boston has been able to leverage machine learning effectively in the short term with a focus on being a \u201cclever,\u201d not a \u201csmart,\u201d city.[vii]<\/a> This means investing in improving existing infrastructure rather than trying to develop holistic platforms for capturing and monitoring data. For example, Boston recently used machine learning to improve its existing bus tracking app, leading to a 10% increase in accuracy.[viii]<\/a> In the medium term, Boston is focused on two main initiatives: collecting additional data and improving the efficiency of broader city services. For example, Boston launched Vision Zero, which allows residents to text safety concerns on city roads to a hotline. The goal is to use the data alongside other information like road conditions and traffic to predict vehicle crashes. Another initiative involves building out an algorithm that can use restaurant registration information to predict health code violations and improve inspection findings by 20-25%.[ix]<\/a><\/p>\n While Boston has made significant progress in machine learning, I believe they could do more. In the short term, Boston should expand its private-public partnership by incentivizing private companies to use machine learning in ways that impact citizens. For example, a South Korean city was able to help companies cut building operating costs by 30% using smart sensors to regulate water and electricity usage.[x]<\/a> Boston could provide tax incentives to real estate companies that develop these technologies.<\/p>\n In the medium term, I recommend Boston focus on two main things: improving data collection and working towards a smart city platform. In general, lot of collected data, estimated around 99.5%, is unlabeled.[xi]<\/a> Cities like Boston need to think about how to best capture usable data. One potential way is through additional crowd-sourcing from citizens. Second, Boston should work towards a smart city platform that incorporates sensors into a holistic solution for the city. One idea is to pilot a test neighborhood that is fully \u201csmart.\u201d LStar Ventures is building a completely connected mini-city in Weymouth Massachusetts, which could serve as a potential partnership for the city.[xii]<\/a><\/p>\n Going forward, there are two main questions Boston needs to consider. First, how should they leverage private partnerships, and will those partnerships potentially violate citizen\u2019s trust on how data is handled? Second, how should Boston think about making its investments in machine learning (i.e., tactically solving specific problems as currently done or through broader long-term platform solutions)?<\/p>\n Word Count = 748 words<\/p>\n [i]<\/a> Frost & Sullivan Global 360 Research Team. \u201cSmart City Scorecard: Top Global Smart Cities Maturing by Building on Foundational Areas of Excellence.\u201d Frost & Sullivan.<\/em> 15 December 2017.<\/p>\n [ii]<\/a> Thibodeaux, Todd. \u201cSmart Cities are Going to be a Security Nightmare.\u201d 性视界 Business Review.<\/em> 28 April 2017.<\/p>\n [iii]<\/a> Jay, Jonathan. \u201cCan Algorithms Predict House Fires?\u201d 性视界 Kennedy School Ash Center: Data-Smart City Solutions.<\/em> 6 March 2017. https:\/\/datasmart.ash.harvard.edu\/news\/article\/can-algorithms-predict-house-fires-990<\/a>. Accessed November 12, 2018.<\/p>\n [iv]<\/a> Bousquet, Chris. \u201cCan Better Data Make Zero Traffic Deaths a Reality?\u201d 性视界 Kennedy School Ash Center: Data-Smart City Solutions. <\/em>10 October 1017. https:\/\/datasmart.ash.harvard.edu\/news\/article\/can-better-data-make-zero-traffic-deaths-a-reality-1138<\/a>. Accessed November 12, 2018.<\/p>\n [v]<\/a> R. W. Thomas and J. M. Vidal, “Toward detecting accidents with already available passive traffic information,” 2017 IEEE 7th Annual Computing and Communication Workshop and Conference (CCWC), Las Vegas, NV, 2017, pp. 1-4.<\/p>\n [vi]<\/a> Woetzel, Jonathan, Remes, Jaana, et al. \u201cSmart cities: Digital solutions for a more livable future.\u201d McKinsey Global Institute.<\/em> June 2018.<\/p>\n [vii]<\/a> City of Boston Office of New Urban Mechanics. \u201cBoston Smart City Plabook.\u201d https:\/\/monum.github.io\/playbook\/<\/a>. Accessed November 12, 2018.<\/p>\n [viii]<\/a> MBTA. \u201cBetter Bus Tracking and Predictions.\u201d 11 October 2018. https:\/\/www.mbta.com\/projects\/better-bus-tracking-and-predictions<\/a>. Accessed November 12, 2018.<\/p>\n [ix]<\/a> Hillenbrand, Katherine. \u201cCase Study: Boston\u2019s Citywide Analytics Team.\u201d 性视界 Kennedy School Ash Center: Data-Smart City Solutions.<\/em> 15 May 2017. https:\/\/datasmart.ash.harvard.edu\/news\/article\/case-study-bostons-citywide-analytics-team-1043<\/a>. Accessed November 12, 2018.<\/p>\n [x]<\/a> Thibodeaux, Todd. \u201cSmart Cities are Going to be a Security Nightmare.\u201d 性视界 Business Review.<\/em> 28 April 2017.<\/p>\n