Search

The analysis of human behavior is the basis of understanding many social phenomena. Based on a large database of taxi billing system, this paper provides an analysis of human mobility data in an urban area of using taxi services in Shanghai. By studying the spatial temporal data of taxi services, it shows that the distribution of running time interval is approximated by Power-law distribution. And it also reflects that the taxi services may have the short-time and small-scaled characteristics, which seems to be consistent with human travel patterns.

Classic genetic algorithm is not suitable to solve traveling salesman problem, because the encoding of the traveling salesman problem is either the permutations of the cities or the combinations of edges, which can not be directly operated by ordinary crossover or mutation operators. This paper introduces a novel encoding for traveling salesman problem, which can be operated by the ordinary crossover and mutation operators; and in order to accelerate the rate of the convergence, a novel local improvement approach is also presented.

TSP is a well-known NP-hard problem. Although many algorithms for solving TSP, such as linear programming, dynamic programming, genetic algorithm, anneal algorithm, and ACO algorithm have been proven to be effective, they are not so suitable for the more complicated large scale TSP. This paper offers a method to decompose the large-scale data into several small-scale data sets by its relativity; and the results of each small-scale data set which represents a small-scale TSP compose the whole result of the large-scale TSP. An adaptive clustering method is presented and a novel genetic algorithm for TSP is described in this paper.

Money laundering (ML) involves moving illicit funds, which may be linked to drug trafficking or organized crime, through a series of transactions or accounts to disguise origin or ownership. China is facing severe challenge on money laundering with an estimated 200 billion RMB laundered annually. Decision tree method is used in this paper to create the determination rules of the money laundering risk by customer profiles of a commercial bank in China. A sample of twenty-eight customers with four attributes is used to induced and validate a decision tree method. The result indicates the effectiveness of decision tree in generating AML rules from companies' customer profiles. The anti-money laundering system in small and middle commerical bank in China is highly needed.

Support vector machine (SVM), which is based on statistical learning theory (SLT), has shown much better performance than most other existing machine learning methods, which are based on the traditional statistics. The original SVM was developed to solve the dichotomy classification problem. Various approaches have been presented to solve multi-class problems. Using multi-class SVM classifier we have obtained high class rate of 95.4% in remote sensing image classification. However for the class number of remote sensing image is much great, manually obtaining of training samples is a much time-consuming work. Hence, we present a multi-class SVM based semi-supervised approach. We choose the initial cluster centers manually first, then label the samples as the training ones automatically with fuzzy C-means clustering algorithm. It is believed that this method upgrades the classification efficiency greatly with practicable class rate.

A framework for consciousness is introduced based upon traditional artificial neural network models. This astral framework reflects explicit connections between two parts of brain: one global working memory and distributed modular cerebral networks related to specific brain functions. Accordingly, this framework is composed of two layers, physical mnemonic layer and abstract thinking layer, which cooperate together to accomplish information storage and cognition using algorithms of how these interactions contribute to consciousness: (1) the bottom-up-projection process whereby cerebral subsystems group distributed signals into coherent object patterns; (2) the top-down-attention process whereby global workspace selectively responses to highly impressive patterns; and (3) the horizontal resonant learning process whereby global workspace steadily adjusts its network to adapt to environmental changes. Using this framework, various sorts of human brain actions can be explained, leading to a general approach for analyzing brain functions.

Efficient application scheduling is critical for achieving high performance in heterogeneous computing systems (HCS). Because an application can be partitioned into a group of tasks and represented as a directed acyclic graph (DAG), the problem can be stated as finding a schedule for a DAG to be executed in a HCS so that the schedule length can be minimized. The tasks scheduling problem is NP-hard in general, except in a few simplified situations. In order to obtain optimal or suboptimal solutions, a large number of scheduling heuristics have been presented in the literature. Genetic algorithm (GA), as a power tool to achieve global optimal, has been successfully used in this field. This work presents a new hybrid genetic algorithm to solve the scheduling problem in HCS. It uses a direct method to encode a solution into chromosome. Topological sort of DAG is used to repair the offspring in order to avoid yielding illegal or infeasible solutions, and it also guarantees that all feasible solutions can be reached with some probability. In order to remedy the GA's weakness in fine-tuning, this paper uses a greedy strategy to improve the fitness of the individuals in crossover operator, based on Lamarckian theory in the evolution. The simulation results comparing with a typical genetic algorithm and a typical list heuristic, both from the literature, show that this algorithm produces better results in terms of both quality of solution and convergence speed.

Geospatial computing today is highly manual requiring large amounts of human capital. The systems used for processing were designed for stand-alone environments, are inappropriate for modern distributed, heterogeneous network environments as Internet due to their closed architecture, and their lack of interoperability, reusability, and flexibility. The need of an intelligent architecture that can operate in complicated mapping situations and new/unknown environments is becoming manifest. Intelligent agent is the rational entity with big granularity, high-intelligence and autonomy characteristics, and Multi-agent system (MAS) is composed of a number of agents capable of communicating, coordinating, cooperating and even competing with each other. It provides a new approach to implement large-scale, distributed, self-adaptable and complicated software system. In this paper, a multi-agent framework is presented to facilitate constructing intelligent distributed service-oriented geospatial computing, which had been partly applied in a real project successfully.

The tasks scheduling problem is a key factor for a parallel multiprocessor system to gain better performance. Because a task can be partitioned into a group of subtasks and represented as a DAG (directed acyclic graph), so the problem can be stated as finding a schedule for a DAG to be executed in a parallel multiprocessor system so that the schedule length can be minimized. The tasks scheduling problem is NP-hard in general, except in a few simplified situations. In order to obtain optimal or suboptimal solutions, a large number of scheduling heuristics have been presented in the literature. The most studied heuristics are based on list heuristic. Recently, genetic algorithm has received much attention as a class of heuristic. This paper presents a new genetic algorithm called TEOL (task execution order list based genetic algorithm) to solve the scheduling problem in parallel multiprocessor systems. It guarantees that all feasible schedules can be reached with some probability, and because the TEOL is based on the restrain of predecessor relationship of the DAG only, so other heuristics can be combined into it to improve the performance. Simulation results comparing with two genetic algorithms and a list algorithm, both from the literature, show that TEOL produces encouraging results in terms of quality of solution and execution speed.

The focus of Computer-Aided Plant-Identification (CAPI) is the stable features extraction of plant, such as sawtooth number of leaf edge for some species of plants. Because the shape of the sawteeth varies greatly, they cannot be depicted in rigid mathematic method. However, a trained SVM (Support Vector Machine) with good adaptability can be applied to classify sawtooth and nonsawtooth samples. The samples can be obtained by a rectangular sample window sliding along the edge of the leaf, and then be rotated to a standard pose for decreasing the complexity of identification. By avoiding repeated sampling and counting of the same sawtooth, the algorithm presented in the paper accomplishes automatic counting of the sawtooth number. The results of the experiment show that the SVM-based method works well.