The Relationship between Decision Trees and the Scale of Train Data Set

Decision Trees need train samples in the train data set to get classification rules. If the number of train data was too small, the important information might be missed and thus the model could not explain the classification rules of data. While it is not affirmative that large scale of train data set can get well model. This Paper analysis the relationship between decision trees and the train data scale. We use nine decision tree algorithms to experiment the accuracy, complexity and robustness of decision tree algorithms. Some results are demonstrated.

Fiber Dispersion and Nonlinearity Influences on Transmissions of AM and FM Data Modulation Signals in Radio-Over-Fiber System

Transmission properties of data amplitude modulation (AM) and frequency modulation (FM) in radio-over-fiber (RoF) system are studied numerically. The influences of fiber dispersion and nonlinearity on different microwave modulation schemes, including double side band (DSB), single side band (SSB) and optical carrier suppression (OCS), are investigated and compared. The power penalties at the base station (BS) and the eye opening penalties of the recovered data at the end users are both calculated and analyzed. Numerical simulation results reveal that the power penalty of FM can be drastically decreased due to the larger modulation depth it can achieve than that of AM. The local spectrum broadening around subcarrier microwave frequency of AM due to fiber nonlinearity can also be eliminated with FM. It is demonstrated for the first time that the eye openings of the FM recovered data can be controlled by its modulation depths and the coding formats. Negative voltage encoding format was used to further decrease the RF frequency thus increase the fluctuation period considering their inverse relationship.

Web-based spatiotemporal visualization of marine environment data

With long-term marine surveys and research, and especially with the development of new marine environment monitoring technologies, prodigious amounts of complex marine environmental data are generated, and continuously increase rapidly. Features of these data include massive volume, widespread distribution, multiple-sources, heterogeneous, multi-dimensional and dynamic in structure and time. The present study recommends an integrative visualization solution for these data, to enhance the visual display of data and data archives, and to develop a joint use of these data distributed among different organizations or communities. This study also analyses the web services technologies and defines the concept of the marine information gird, then focuses on the spatiotemporal visualization method and proposes a process-oriented spatiotemporal visualization method. We discuss how marine environmental data can be organized based on the spatiotemporal visualization method, and how organized data are represented for use with web services and stored in a reusable fashion. In addition, we provide an original visualization architecture that is integrative and based on the explored technologies. In the end, we propose a prototype system of marine environmental data of the South China Sea for visualizations of Argo floats, sea surface temperature fields, sea current fields, salinity, in-situ investigation data, and ocean stations. An integration visualization architecture is illustrated on the prototype system, which highlights the process-oriented temporal visualization method and demonstrates the benefit of the architecture and the methods described in this study.