Anton von Schantz

Anton von Schantz

Data Science Consultant | ML Engineer

Aalto University

About

I am keen on helping businesses make better decisions by combining their data and machine learning and modeling expertise. I am an analytics consultant with a decade of industry experience. I also have a scientific background in systems and operations research (I am defending my Ph.D. thesis in June 2021).

Visit also my crowddynamics and evacuation research website crowddynamics.aalto.fi.

Tech stack: Python, SQL, R, JavaScript, Flask, Azure, Docker, Tensorflow, Gurobi

Education

  • PhD in Systems and Operations Research, 2021

    Aalto University

  • MSc in Systems and Operations Research, 2014

    Aalto University

  • BSc in Systems and Operations Research, 2014

    Aalto University

Experience

 
 
 
 
 

Doctoral Researcher

Systems Analysis Laboratory, Aalto University

Oct 2014 – Present Espoo
  • I developed novel numerical simulation models and optimization heuristics for human crowd evacuations. The simulations were run on Aalto University supercomputer.
  • My research was featured in the Finnish periodical magazine “Tekniikan Maailma”.
  • The research resulted in 6 publications, and it was done in collaboration with VTT Fire Safety Research Group and Aalto University Department of Civil Engineering. I also personally applied for and received grants from the Foundation for Aalto University Science and Technology, and the Finnish Science Foundation for Technology and Economics.
  • I managed 3 research assistants whose role was software development.
  • I taught 21 university level courses on artificial intelligence and mathematical optimization and received the teacher of the year award 3 times.
 
 
 
 
 

Freelance Data Scientist

von Schantz Consulting

Oct 2020 – Mar 2020
  • The client was a large Finnish construction company.
  • A proof-of-concept project to identify the main demand factors for their new developments. Together with the client we developed a machine learning model for demand prediction.
  • See my profile at https://www.rootsof.ai/services-for-data-professionals
 
 
 
 
 

Secretary

Finnish Operations Research Society

Sep 2015 – Jan 2017 Helsinki
  • I promoted operations research and its applications by organizing and marketing 4 events on using different operations research methods in business and other practical applications.
  • I was the editor of 3 issues of the society’s journal.
 
 
 
 
 

Data Scientist Trainee

Comptel Oy (currently owned by Nokia)

Jun 2014 – Present Helsinki
  • I developed an interactive visualization tool for their analytics product.
  • The tool visualizes important teleoperator company customer data, and acts as an aid for decision-making.
 
 
 
 
 

Research Assistant

Systems Analysis Laboratory, Aalto University

Jun 2012 – Aug 2013
  • As an undergraduate, I developed a novel computationally light evacuation model that can be used for real-time evacuation simulations.
  • In collaboration with VTT Fire Safety Research Group, I programmed a model specially suited to describe crowd turbulence, i.e., the main phenomenon behind crowd disasters.
 
 
 
 
 

Data Scientist

Department of Marketing, Aalto University

May 2011 – May 2012
  • I worked in a marketing research project done for a large Finnish food company.
  • I analyzed and identified the effect of the company’s and its competitors marketing activities on the food company’s sales.
  • Based on historical sales data I developed a sales forecast model for different food products.
 
 
 
 
 

Research Assistant

Risk Management Unit, Helsinki City Rescue Department

Jul 2009 – Sep 2009
  • I developed a statistical model for accident rates in Finland that was used in a larger project for reallocation of emergency resources in Finland.
  • In the project, I collaborated with VTT Fire Safety Research Group and Emergency Services Academy of Finland.

Publications

My research from 2014 to 2020. First, the focus is on mathematical and physical modeling of the dynamics of an evacuating crowd. Later, optimization methods and heuristics like stochastic optimization, portfolio optimization and genetic algorithms are applied to develop tools for evacuation planners and risk management.

Cellular automaton evacuation model coupled with a spatial game (2014)

Cellular automaton evacuation model coupled with a spatial game (2014)

My first research paper concerns the modeling of an evacuation through a bottleneck with a computationally light cellular automaton (CA) model. In it, the evacuating crowd is modeled as a multi-agent system, and the agents move in a discrete square grid according to probabilistic rules. Their decision-making is modeled with a game-theoretical model. The game is played locally and the strategy choice affects the movement of the agents.

Spatial game in cellular automaton evacuation model (2015)

Spatial game in cellular automaton evacuation model (2015)

We perform extensive numerical calculations with the model from (von Schantz & Ehtamo, 2014). We show that most collective effects observed in crowds evacuating through bottlenecks can be produced with the simple local-decision making model.

Twotype multiagent game for egress congestion (2017)

Twotype multiagent game for egress congestion (2017)

The model from (von Schantz & Ehtamo, 2014) is extended to include agents with different risk perceptions. In the numerical simulations with the model we show that agents evacute in irregular successions. The more threatening they perceive the situation, the longer the time lapse on average between two consecutively evacuated agents is.

Pushing and overtaking in a spatial game of exit congestion (2019)

Pushing and overtaking in a spatial game of exit congestion (2019)

The crowd dynamics in an exit congestion is studied with a physically realistic evacuation model. The same game-theoretical model used already in (von Schantz & Ehtamo, 2014) is coupled to it. The physics of an evacuating crowd are explained starting from the microscopic physical interactions and individual decision-making of the agents.

Minimizing the evacuation time of a crowd from a complex building using rescue guides (2020)

Minimizing the evacuation time of a crowd from a complex building using rescue guides (2020)

The previous publications considered the realistic and computationally efficient modeling crowd dynamics. In this manuscript, an optimization procedure for solving the minimum time evacuation plan using rescue guides is presented. The method is applied on an evacuation from a conference building, which gives an evacuation plan that decreases the evacuation time by 80%. A high-performance computing cluster is used to solve the problem.

Minimization of mean-CVaR evacuation time of a crowd using rescue guides: a scenario-based approach (2020)

Minimization of mean-CVaR evacuation time of a crowd using rescue guides: a scenario-based approach (2020)

Methods from investment portfolio optimization is used to solve the minimum time evacuation using rescue guides. With enough rescue guides the evacuation time in all scenarios can be optimized. If there is a restraint on the number of guides, there is a tradeoff between average- and worst case-performing evacuation plans.