Peter C. Nelson

Evolving accurate and compact classification rules with gene expression programming

Classification is one of the fundamental tasks of data mining. Most rule induction and decision tree algorithms perform a local, greedy search to generate classification rules that are often more complex than necessary. Evolutionary algorithms for pattern classification have recently received increased attention because they can perform global searches. In this paper, we propose a new approach for discovering classification rules by using gene expression programming (GEP), a new technique of genetic programming (GP) with linear representation. The antecedent of discovered rules may involve many different combinations of attributes. To guide the search process, we suggest a fitness function considering both the rule consistency gain and completeness. A multiclass classification problem is formulated as multiple two-class problems by using the one-against-all learning method. The covering strategy is applied to learn multiple rules if applicable for each class. Compact rule sets are subsequently evolved using a two-phase pruning method based on the minimum description length (MDL) principle and the integration theory. Our approach is also noise tolerant and able to deal with both numeric and nominal attributes. Experiments with several benchmark data sets have shown up to 20% improvement in validation accuracy, compared with C4.5 algorithms. Furthermore, the proposed GEP approach is more efficient and tends to generate shorter solutions compared with canonical tree-based GP classifiers.

An automated GPS-based prompted recall survey with learning algorithms

Abstract Using GPS technology in the collection of household travel data has been gaining importance as the technology matures. This paper documents recent developments in the field of GPS travel surveying and ways in which GPS has been incorporated into or even replaced traditional household travel survey methods. A new household activity survey is presented which uses automated data reduction methods to determine activity and travel locations based on a series of heuristics developed from land-use data and travel characteristics. The algorithms are used in an internet-based prompted recall survey which utilizes advanced learning algorithms to reduce the burden placed on survey respondents. Initial results of a small pilot study are discussed and potential areas of future work are presented.

Perimeter search

A technique for improving heuristic search efficiency is presented. This admissible technique is referred to as perimeter search since it relies on a perimeter of nodes around the goal. The perimeter search technique works as follows: First, the perimeter is generated by a breadth-first search from the goal to all nodes at a given depth d. The path back to the goal along with each perimeter nodes state descriptor are stored in a table. The search then proceeds normally from the start state. During each node generation, however, the current node is compared to each node on the perimeter. If a match is found, the search can terminate with the path being formed with the path from the start to the perimeter node together with the previously stored path from the perimeter node to the goal. Both analytical and experimental results are presented to show that perimeter search is more efficient than IDA∗ and A∗ in terms of time complexity and number of nodes expanded for two problem domains.

Optimization of high-speed multistation SMT placement machines using evolutionary algorithms

Surface mount technology (SMT) is a robust methodology that has been widely used in the past decade to produce circuit boards. Analyses of the SMT assembly line have shown that the automated placement machine is often the bottleneck, regardless of the arrangement of these machines (parallel or sequential) in the assembly line. Improving and automating the placement machine is a key issue for increasing SMT production line throughput. This paper presents experimental results using genetic algorithms to optimize the feeder slot assignment problem for a high-speed parallel, multistation SMT placement machine. Four crossover operators, four selection methods, and two probability settings are used in our experiments. A penalty function is used to handle constraints. A comparison of genetic algorithms with several other optimization methods (human experts, vendor supplied software, expert systems, and local search) is presented, which supports the use of genetic algorithms for this problem.

The Intelligent Travel Assistant

The ultimate goal of the Intelligent Travel Assistant (ITA) is the fusion and development of a number of technologies into a device that will increase the efficiency of our transportation network through increased use of mass transit and ride sharing. The ITA will include: a) dynamic ridesharing, b) spatio-temporal database management, c) wireless communications, and d) a management framework. The ITA itself is envisioned to be a compact portable device with the capability to plan multi-modal routes for its user. The traveler will enter a desired destination into the ITA and it will formulate several plans to got the user to this destination. The ITA will make use of wireless Internet technology to send and receive traffic information, transit schedules, and arrange for payment of fares. The Global Positioning System (GPS) will be used to track the users current position for use in route planning and also its it means of assessing traffic conditions while en-route. A spatio-temporal database management system will be used to efficiently track ITA positions in real-time.

Real-Time Street Parking Availability Estimation

Real-time parking availability information is important in urban areas, and if available could reduce congestion, pollution, and gas consumption. In this paper, we present a software solution called PhonePark for detecting the availability of on-street parking spaces. The solution uses the GPS and/or accelerometer sensors in a travelers mobile phone to automatically detect when and where the traveler parked her car, and when she released a parking slot. PhonePark can also utilize the mobile phones Bluetooth sensor or piggyback on street parking payment transactions for parking activity detection. Thus, the solution considers only mobile phones and does not rely on any external sensors such as cameras, wireless sensors embedded in the pavements, or ultrasonic sensors on vehicles. Further contributions include an algorithm to compute the historical parking availability profile for an arbitrary street block and algorithms to estimate the parking availability in real-time for a given street block. The algorithms are evaluated using real-time and real world street parking data.

Cluster-Based framework in vehicular ad-hoc networks

The application of Mobile Ad Hoc Network (MANET) technologies in the service of Intelligent Transportation Systems (ITS) has brought new challenges in maintaining communication clusters of network members for long time durations. Stable clustering methods reduce the overhead of communication relay in MANETs and provide for a more efficient hierarchical network topology. During creation of VANET clusters, each vehicle chooses a head vehicle to follow. The average number of cluster head changes per vehicle measures cluster stability in these simulations during the simulation. In this paper we analyze the effect of weighting two well-known clustering methods with the vehicle-specific position and velocity clustering logic to improve cluster stability over the simulation time.

Traffic model for clustering algorithms in vehicular ad-hoc networks

Inter-vehicle communication by means of wireless Ad Hoc networking has the potential to improve traffic safety and comfort tremendously. Therefore, the application of Vehicular Ad Hoc Networks (VANETs) in the service of Intelligent Transportation Systems (ITS) has been highly focused in recent years. Derived from the successful outcome of a cluster-based framework in Mobile Ad Hoc Networks (MANETs), we apply this network topology to VANETs. Unfortunately, previous studies lack realistic modeling of vehicle mobility and evaluation of clustering performance so they may not correlate well with performance in a real deployment. Hence, in this paper, we propose a realistic micro- simulation model with the hope of contributing to clustering research in VANETs, and demonstrate how clustering algorithms work on it.

Optimization of high-mix printed circuit card assembly using genetic algorithms

The purpose of this paper is to present an overview of the factors affecting the cycle time of printed circuit card assembly (PCCA) in high-mix environments and demonstrate a technique for improving machine throughput. We have concentrated our research on optimizing the portion of the PCCA manufacturing process performed by high-speed placement machines (chip shooters). A crucial factor affecting the throughput of a chip shooter is the assignment of components to the feeder slots. Genetic algorithms were employed to find a near optimal assignment of the feeder carriage. Results for various genetic operators in this problem domain are presented.

A predicate-transition net model for multiple agent planning

Abstract When multiple agents are available, some actions can be performed concurrently. Performing actions concurrently can be conveniently modeled with a predicate-transiton net ( pr t net). We provide an algorithm to construct a pr t net model for a robot planning system represented in a STRIPS-like specification language. By applying our linear search algorithm to the pr t net model, we can obtain a plan in which parallel actions are explicitly shown. The idea behind our search algorithm is based on a combination of bidirectional search and means-ends analysis. Our search strategy expands fewer nodes than other search algorithms. For example, it expands only necessary nodes in solving Sussmans anomalous problem.