Huailin Dong

Research and improvement of ant colony algorithm based on TSP

Ant colony algorithm is a new algorithm of heuristic bionic calculation. Now, it has been widely applied in many fields of combinatorial optimization. This paper elaborates the basic principle and mathematical model of typical ant colony algorithm for solving the traveling salesman problem, and analyzes impact of the optimal parameters to the performance of algorithm. Based on its shortages, an improved algorithm by dynamically adjusting parameters is proposed. Finally, the paper gives the simulation result, and it indicates the improved algorithm has a better performance.

Development and Application of Mobile Nursing Information System Based on PDAs

Combined with the technology of PDA and WLAN, and in view of limitations of wired networking in current hospital nursing workstations, this paper designs and implements a mobile nursing information system based on B/S taking PDA as its running platform. This system is a natural extension of the existing HIS, which takes patients as its center. And its application can optimize the work flow of clinical nursing, raise work efficiency of nursing personnel, promote service quality of medical treatment, improve hospital nursing management, and reduce medical accidents.

A method for automatically determining The number of clusters of LAC

The algorithm of locally adaptive clustering for high dimensional data (LAC) processes soft subspace clustering by local weightings of features. To solve the localization of LAC in specifying the number of clusters, this paper reworks the validity index for fuzzy clustering to evaluate the clustering results of LAC. Compared with real clustered data, the method is proved feasible. In the new algorithm, validity function is calculated under different clusters to discover the best clustering number. Experiments have shown that the improved LAC could search for the true number of clusters in high dimensional data sets automatically, as well as elevation of its clustering accuracy.

Mobile Guardian: A novel positioning and monitoring system for outdoor special users based on GPS

Different to traditional vehicle positioning and navigation systems, requirements of positioning operations for individuals are always contingently, and it pays more attention to making navigation devices portable and easy to use. Based on GPS and GSM, a novel positioning and monitoring system for outdoor special users called Mobile guardian is presented in this paper. Through analyzing the structure of GPS data, location-related information such as longitude and latitude of users can be extracted from GPS data stream. The GSM module which supported AT commands is used to transmit these location data and telecommands between users and the monitoring center. And the technology for encapsulating and parsing XML spacial data is utilized for Google Earth to display the detailed geographical information on the screen. Experiments show that the system is effective and could be popularized to family monitoring.

Research on statistics-based model for E-commerce user purchase prediction

This paper describes our work for ALIDATA DISCOVERY competition. Through analyzing massive real-world user action data provided by Tmall, one of the largest B2C online retail platforms in China, we try to predict future user purchases. The prediction results are judged by F1 Score that is consist of two parts, precision and recall rate. The provided data set contains more than 500 million action records from over 12 million distinct users. Such a massive data set drives us to finish the task in MapReduce fashion on the Open Data Processing Service (ODPS) platform. According to statistical results, we classify all users into different groups firstly. Then the rule model, timing model, statistics model are adopted for predicting future user purchases. By comparison, the statistics model obtains the best F1Score.

Large-scale text clustering based on improved K-means algorithm in the storm platform

With the web information dramatically increases, Distributed processing of mass data through a cluster have been the focus of research field. An efficient distributed algorithm is the determinant of the scalability and performance in data analyses. This dissertation firstly studies the operation mechanism of Storm, which is a simplified distributed and real-time computation platform. Based on the Storm platform, an improved K-Means algorithm which could be used for data intensive computing is designed and implemented. Finally, the experience results show that the K-Means clustering algorithm base on Storm platform could obtain a higher performance in experience and improve the effectiveness and accuracy in large-scale text clustering.

Research and design of digital museum based on virtual reality

Digital museum is an important trend in the development of museums in recent years, and virtual reality technology in the digital museum is one of the most important issues in museum construction and it has broad development prospects. This paper aims to research into two kinds of key virtual reality technologies: image-based 360 degree panoramic display and 3D modeling technology, exploring various kinds of virtual museums based on these two technologies. It discusses the advantages and limitations of the various kinds of VDMs and proposes the design of the Virtual Museum based on 360 - degree panorama, web 3D technology and 3D modeling technology. And its application in the design of Xiamen Overseas Chinese Museum has obtained good results.

Research on predicted model of least squares support vector machine based on genetic algorithm

There is huge amount of data with complex uncertainty in the stock market. Meanwhile, efficient stock prediction is important in financial investment. This paper puts forward a classified and predicted model based on least squares support vector machine (LS-SVM) in the background of stock investment. This model preprocesses the input vector of stock indexes using the method of Wilcoxon symbols test and factor analysis, and determines the parameter of LS-SVM based on the genetic algorithm, after that classifies the stocks based on growth rate, then is trained using the stock sample. At last this paper verifies the model with the samples. It also presents a demo to predict the increasing trend of the stock. The result shows that this model owns favorable predicted ability with high correct classification rate.

Research on emotion modeling based on three-dimension emotional space

Combined with the particle system and active field, this paper establishes an artificial emotion model on the basis of the three-dimension emotional space and OCC model. The model converts the external stimuli into the active field in the space, and the potential energy generated by the active field has the combined impact on the particles’ state of motion. The states of motion in the particle system represent the model’s emotional state. In addition, the model outputs a time-varying multi-dimensional vector, and the vector values in all dimensions describe the activation level of the corresponding emotion.

Naive Bayes Classification Algorithm Based on Optimized Training Data

Naive Bayes classification algorithm is an effective simple classification algorithm. Most researches in traditional Naive Bayes classification focus on the improvement of the classification algorithm, ignoring the selection of training data which has a great effect on the performance of classifier. And so a method is proposed to optimize the selection of training data in this paper. Adopting this method, the noisy instances in training data are eliminated by user-defined effectiveness threshold, improving the performance of classifier. Experimental results on large-scale data show that our approach significantly outperforms the baseline classifier.