【摘 要】
：
To improve the training speed of traditional support vector regression(SVR)algorithm on large datasets,an adaptive resample strategy is proposed to reduce the training dataset without loss of performa
【机 构】
：
Electrical Engineering College,Henan University of Science and Technology,Luoyang 471023
论文部分内容阅读
To improve the training speed of traditional support vector regression(SVR)algorithm on large datasets,an adaptive resample strategy is proposed to reduce the training dataset without loss of performances.As for most samples collected from physical world,the age of the data does affect the contribution on predicting the coming status.A resample strategy is embedded into the training procedure of SVR according to their timeliness.First,the samples are divided into several groups according to their age from current time.Then,the resample numbers are gained by Gaussian process according to the average age of each group,the resampled instances are used as training and testing sets.By this way,an improved regression algorithm is constructed for large datasets with time sensitive problems.The simulation results on several benchmark datasets show that the improved algorithm embedded adaptive resampling has much advantage in the training speed than traditional and random sampling regression algorithm.
其他文献
针对3D 打印设备、3D 相关设备和运行数据的统一建模问题，提出并建立3D 网络系统合成模型，包含数据合成类型定义模型、单纯数据类型定义模型、组合数据类型定义模型、数据合成约束系统，实现3D 打印设备与3D 相关设备间的互操作、信息共享、协同工作；建立3DCNS 和3DNSC 的配套通讯体系，构建网络与设备的综合信息模型和对应通讯机制，实现网络与设备的自我表述。通过集成禁用品保护、在线防闭锁、区域
Industrial process plants are often characterized with problems of variable drifts and change of state.The time difference(TD)model was proposed by researchers to handle the drifting problems.However,
The stretching process plays a key role in the quality of the final fiber product.Due to the fiber stretching process with inborn nonlinearity,a global nonlinear model can be achieved by the way of th
This paper focuses on the batch to batch temperature control of a semi-batch chemical reactor with heating/cooling system.Without extensive modeling investigations,a 2D general predictive iterative le
This paper addresses the problem of input and state joint estimation for linear discrete-time stochastic systems without direct feedthrough from unknown input to output.The input estimate is obtained
This paper develops a multivariable multiple-model adaptive control scheme for adaptive state feedback state tracking control of systems whose plant-model matching conditions are uncertain and paramet
Since the output voltage of each DG is different due to the effects of mismatched line impedance,the reactive power could not be shared accurately with conventional droop control.This paper proposes t
The reconstruction method,which defined in the frame of principal component analysis,is a popular and representative method for fault diagnosis.This paper analyzes the ability of the reconstruction me
A New cost estimation method based on Monte-Carlo simulation and market investigation is proposed to consider both the uncertainties of cost estimation and their market drivers.There are several criti
Focused on the problem that the take-off of a small unmanned helicopter(SUH)mainly depends on the control of the hand-operated,an automatic take-off method for SUH is proposed and the circuit is imple