City:

Syracuse, New York.

Organization:

IBM Corporation.

Project Start Date:

September 2011.

Project End Date:

December 2011. (Published)

References:

Appel, S. U., Botti, D., Jamison, J., Plant, L., Shyr, J. Y., & Varshney, L. R. (2013). Predictive analytics can facilitate proactive property vacancy policies for cities. Technological Forecasting and Social Change,89, 161-173. doi:10.1016/j.techfore.2013.08.028

Baker, C. (2017, July 17). City stuck maintaining vacant houses as Land Bank tightens its belt. Retrieved from https://www.syracuse.com/news/index.ssf/2017/07/city_stuck_with_run-down_houses_as_land_bank_turns_its_back.html

IBM’s Smarter Cities Challenge Syracuse Summary Report (Rep.). (2011, December). Retrieved http://prd-ibm-smarter-cities-challenge.s3.amazonaws.com/applications/syracuse-united-states-summary-2011.pdf 

Knauss, T. (2013, April 16). Syracuse assesses new fees for owners of vacant properties. Retrieved from https://www.syracuse.com/news/index.ssf/2013/04/syracuse_assesses_new_fees_for.html 

Problem:

Background:

With the aim of combating the byproducts of vacant residential properties in the city of Syracuse (which include: criminal activity, increased risk of residential fires & lost tax revenues), the city sought IBM for help in understanding, analyzing and predicting residential vacancy. 

Objective: Develop a proactive approach that would allow the city to prevent property vacancy.

Requirements: 

1)Identify indicators for factors that are contributing to vacancy.

2)Integrate & analyze data from different sources within the city.

3)Develop a predictive model that allows for assessing the impacts of possible courses of action.

Technical Solution:

1.0 Vacant Property Predictive System of Systems with 5 elements:

   1.1 Neighbourhood System

   1.2 Planning & Development System

   1.3 Support Service System

   1.4 Common Service System 

   1.5 Predictive Situational Analysis System

2.0 Predictive Situational Analysis System with 6 components:

    2.1 Data clearinghouse: normalizes data.

    2.2 Prediction: predicts vacancy state & risk using regression, ensembles of decision trees &     

          bipartite ranking.   

    2.3 Cost Estimation: estimates direct and indirect costs of falling into/out of vacancy. 

    2.4 Decision Analysis: analyses impacts of various courses of actions.

    2.5 Event Correlation: provides the prediction component with information regarding actions  

          taken.

    2.6 Dashboard: provides stakeholders with access to results. 

Datasets Used:

• Dataset 1: Parcel Data, Bureau of Planning & Sustainability. 
• Dataset 2: Neighbourhood indicator data, Bureau of Planning & Sustainability (2000 census).                        
• Dataset 3: Police call data for 2011, Syracuse Police Department. 
• Dataset 4: Population race data for 2010, U.S. Census Bureau. 
• Dataset 5: Vacancy data for 2010, U.S. Census Bureau. 
• Dataset 6: Vacancy data, United States Postal Service
• Dataset 7: Vacancy data, Department of Housing & Urban Development. 

Outcome:

IBM's solution has lead to the passing of a legislation that requires owners of vacant 

properties in Syracuse to register with the city and pay an annual fee. As of July 2017, 

however, Syracuse was still dealing with 1,800 vacant properties due to funding 

constraints. 

Issues that arose:

Data from disparate sources required normalizing into standardized forms. 

Status:

Most likely still in use by city but not confirmed.

Entered by:

Solwan Aldeeb, s.aldeeb@mail.utoronto.ca