Operations Research and Optimization: Improving Decisions from Data

Data Science Milan
3 min readJun 10, 2018

--

“Moving towards prescriptive analytics”

On 24th May 2018 at Mikamai, Data Science Milan has organized another interesting meetup about operations research topic. It is the application of scientific methods, techniques and tools used looking for the optimum solutions for the problems.

“Operations Research & Optimization: A New Dimension to Data Science”, by Andrea Taverna, Università degli Studi di Milano

Data science (DS) and operation research (OR) can be seen as complementary to each other, where the first one is more focused on data, how to extract information and knowledge from data to take decisions; the second one tries to evaluate decisions and modelling them in the process with the goal to find optimal solutions. In this way, operations research can be considered as a new dimension of data science.

If you look at the growth of operational research it has a flat trend, meanwhile data science & machine learning have been growing in recent years, but looking at the analytics maturity model by studies from PWC, Gartner and SAS we’ll move towards prescriptive analytics and operation research is positioned in this fourth dimension.

“Optimized Assignment Patterns in Mobile Edge Cloud Networks”, by Alberto Ceselli, Università degli Studi di Milano

Machine learning and operation research can interact in three ways:

-in machine learning there are sub-problems which are optimization problems;

-replacing some heuristic methods with some exacts methods;

-solving prescriptive analytical questions.

An application of prescriptive analytics is developed by Mobile Edge Computing network (MEC): given an existing MEC with virtualization facilities of limited capacity and a set of mobile Access Points (AP) whose data traffic demand changes over time, the aim is to find plans for assigning APs traffic to MEC facilities satisfying each AP demand without exceed MEC facility capacity.

In the data-driven architecture there are two fundamental components: pre-processing and optimization. The first one is used to map the problem and optimization component is used to solve the problem by mathematical programming.

”Optimization modeling in Python”, by Marco Casazza, Università degli Studi di Milano

In the last section has been showed “Pyomo”, a Python module that allows users to formulate optimization problems using Python language.

The first application has regarded knapsack problem: given a set of items, each with a weight and a value, the goal is to determine the number of each item to include in a collection so that the total weight must be less than or equal to a given limit and the total value as large as possible, it can be considered as a maximization profit problem.

The second application has showed flight assignment problem: given a set of flights and a crew of airline company, the goal is to create a weekly plan minimizing the overall cost.

The last example has explained drones surveillance problem: given a number of drones equipped with camera and an area to be controlled, the goal is to optimize the number of drones to cover the whole area.

Written by Claudio G.Giancaterino

Originally published at datasciencemilan.org on June 10, 2018.

--

--

Data Science Milan
Data Science Milan

Written by Data Science Milan

Blog and summary of events of the Data Science Milan community.

No responses yet