一种改进的多源异构数据预处理方法

An Improved Multi-Source Heterogeneous Data Preprocessing Method

数据预处理是所有数据融合中必不可少的内容,该阶段的关键操作是处理噪声数据,去除噪声有助于提升模型的训练结果。基于此,结合已有的研究成果,引入相似度度量的概念,在余弦相似度的基础上,用余弦值表示样本点和聚类中心之间的夹角,用该值对欧氏距离进行加权,将二者结合,构造出一种新的数据去噪方法。

Data preprocessing is an essential step in all data fusion,one of the key operations in this stage is the processing of noisy data,removing noise helps to improve the training results of the model.Based on this,combined with the existing research results,the concept of similarity measure is introduced.On the basis of cosine similarity,the cosine value is used to represent the included angle between the sample point and the cluster center,and the Euclidean distance is weighted with this value.Combining the two,a new data denoising method is constructed.