Stephan Meisel

Synergies of Operations Research and Data Mining

In this contribution we identify the synergies of Operations Research and Data Mining. Synergies can be achieved by integration of optimization techniques into Data Mining and vice versa. In particular, we define three classes of synergies and illustrate each of them by examples. The classification is based on a generic description of aims, preconditions as well as process models of Operations Research and Data Mining. It serves as a framework for the assessment of approaches at the intersection of the two procedures.

Simulated annealing in the presence of noise

Abstract In many practical optimization problems, evaluation of a solution is subject to noise, e.g., due to stochastic simulations or measuring errors. Therefore, heuristics are needed that are capable of handling such noise. This paper first reviews the state-of-the-art in applying simulated annealing to noisy optimization problems. Then, two new algorithmic variants are proposed: an improved version of stochastic annealing that allows for arbitrary annealing schedules, and a new approach called simulated annealing in noisy environments (SANE). The latter integrates ideas from statistical sequential selection in order to reduce the number of samples required for making an acceptance decision with sufficient statistical confidence. Finally, SANE is shown to significantly outperform other state-of-the-art simulated annealing techniques on a stochastic travelling salesperson problem.

Floating Car Data Based Analysis of Urban Travel Times for the Provision of Traffic Quality

The management of urban traffic systems demands information for the real-time control of traffic flows as well as for strategic traffic management. In this context, state-of-the-art traffic information systems are mainly used to control varying traffic flows and to provide collective and individual information about the current traffic situation. However, the provision of information for strategic traffic management as well as for traffic demand dependent planning activities (e.g., in city logistics) is still a potential field of research due to the former lack of reliable city-wide traffic information. Recently, historical traffic data arising from telematics-based data sources provided information for time-dependent route planning, for the improvement of traffic flow models as well as for spatial and time-dependent forecasts. In this chapter, we focus on the analysis of historical traffic data, which serves as a background for sophisticated real-time applications.

Data chain management for planning in city logistics

This contribution is about data chain management enabling route planning in city logistics. The transformation of raw data into reliable decisions requires effective data chain management. The data chain closes the gap between empirical collection of raw traffic data and decision-making in terms of route planning. We define the data chain for the support of route planning in city logistics. The data chain transforms raw empirical traffic data into planning data by first and second level aggregation. The single elements of the data chain are investigated in detail. We discuss basic issues of telematics-based data collection, data cleaning and data integration. The key element of the data chain is the aggregation by cluster analysis. Aggregated data is evaluated by explorative data analysis. Finally, the efficient application of aggregated data for route planning is illustrated.

Tutorial on Stochastic Optimization in Energy—Part I: Modeling and Policies

There is a wide range of problems in energy systems that require making decisions in the presence of different forms of uncertainty. The fields that address sequential, stochastic decision problems lack a standard canonical modeling framework, with fragmented, competing solution strategies. Recognizing that we will never agree on a single notational system, this two-part tutorial proposes a simple, straightforward canonical model (that is most familiar to people with a control theory background), and introduces four fundamental classes of policies which integrate the competing strategies that have been proposed under names such as control theory, dynamic programming, stochastic programming and robust optimization. Part II of the tutorial illustrates the modeling framework using a simple energy storage problem, where we show that, depending on the problem characteristics, each of the four classes of policies may be best.

Tutorial on Stochastic Optimization in Energy—Part II: An Energy Storage Illustration

In Part I of this tutorial, we provided a canonical modeling framework for sequential, stochastic optimization (control) problems. A major feature of this framework is a clear separation of the process of modeling a problem, versus the design of policies to solve the problem. In Part II, we provide additional discussion behind some of the more subtle concepts such as the construction of a state variable. We illustrate the modeling process using an energy storage problem. We then create five variations of this problem designed to bring out the features of the different policies. The first four of these problems demonstrate that each of the four classes of policies is best for particular problem characteristics. The fifth policy is a hybrid that illustrates the ability to combine the strengths of multiple policy classes.

Anticipatory Optimization for Dynamic Decision Making

The availability of todays online information systems rapidly increases the relevance of dynamic decision making within a large number of operational contexts. Whenever a sequence of interdependent decisions occurs, making a single decision raises the need for anticipation of its future impact on the entire decision process. Anticipatory support is needed for a broad variety of dynamic and stochastic decision problems from different operational contexts such as finance, energy management, manufacturing and transportation. Example problems include asset allocation, feed-in of electricity produced by wind power as well as scheduling and routing. All these problems entail a sequence of decisions contributing to an overall goal and taking place in the course of a certain period of time. Each of the decisions is derived by solution of an optimization problem. As a consequence a stochastic and dynamic decision problem resolves into a series of optimization problems to be formulated and solved by anticipation of the remaining decision process.However, actually solving a dynamic decision problem by means of approximate dynamic programming still is a major scientific challenge. Most of the work done so far is devoted to problems allowing for formulation of the underlying optimization problems as linear programs. Problem domains like scheduling and routing, where linear programming typically does not produce a significant benefit for problem solving, have not been considered so far. Therefore, the industry demand for dynamic scheduling and routing is still predominantly satisfied by purely heuristic approaches to anticipatory decision making. Although this may work well for certain dynamic decision problems, these approaches lack transferability of findings to other, related problems.This book has serves two major purposes: It provides a comprehensive and unique view of anticipatory optimization for dynamic decision making. It fully integrates Markov decision processes, dynamic programming, data mining and optimization and introduces a new perspective on approximate dynamic programming. Moreover, the book identifies different degrees of anticipation, enabling an assessment of specific approaches to dynamic decision making. It shows for the first time how to successfully solve a dynamic vehicle routing problem by approximate dynamic programming. It elaborates on every building block required for this kind of approach to dynamic vehicle routing. Thereby the book has a pioneering character and is intended to provide a footing for the dynamic vehicle routing community.

Synergies of Data Mining and Operations Research

Operations research and data mining are consistently described in terms of systems theory. Application targets and process models for both paradigms are compared. Three classes of synergies from the integration of the two are distinguished. This classification can serve as a framework for the assessment of approaches at the intersection of operations research and data mining

Multi-Objective Analysis of Approaches to Dynamic Routing Of a Vehicle

We consider a routing problem for a single vehicle serving customer locations in the course of time. A subset of these customers must necessarily be served, while the complement of this subset contains dynamic customers which request for service over time, and which do not necessarily need to be served. The decision maker’s conflicting goals are serving as many customers as possible as well as minimizing total travel distance. We solve this bi-objective problem with an evolutionary multi-objective algorithm in order to provide an a-posteriori evaluation tool for enabling decision makers to assess the single-objective solution strategies that they actually use in real-time. We present the modifications to be applied to the evolutionary multi-objective algorithm NSGA2 in order to solve the routing problem, we describe a number of real-time single-objective solution strategies, and we finally use the gained efficient trade-off solutions of NSGA2 to exemplarily evaluate the real-time strategies. Our results show that the evolutionary multi-objective approach is well-suited to generate benchmarks for assessing dynamic heuristic strategies. Our findings point into future directions for designing dynamic multi-objective approaches for the vehicle routing problem with time windows.

Serving Multiple Urban Areas with Stochastic Customer Requests

We consider the problem of routing one vehicle for serving customer requests. Customer requests appear randomly over time and must be either confirmed or rejected after becoming known. Our goal is maximization of the number of customer requests served within a given period of time. Customers have different request probabilities and are geographically clustered. The problem reflects a typical situation of logistics service providers covering a number of urban areas with one vehicle. We solve the problem by approximate dynamic programming. Our results are compared to solutions gained from state-of-the-art waiting heuristics.