基于多元统计分析的小样本数据预测模型设计

Design of Prediction Model for Small Sample Data Based on Multivariate Statistical Analysis

若小样本数据预测误差较大,会直接影响数据应用效果,为提升小样本数据预测精度,提出基于多元统计分析的小样本数据预测模型设计方法。将小样本数据放入SPSS软件中,结合自助法完成小样本数据的经验分布分析。基于样本数据经验分布特征,结合具备学习能力的Fisherface算法对小样本上数据实施预分类,建立测试样本类别标签,实现小样本数据的特征提取。通过多元统计分析数据特征的主元成分,确定模型回归函数,结合支持向量机构建数据预测模型,通过上述模型完成小样本数据的精准预测。实验结果表明,使用上述方法开展小样本数据预测时,预测误差较低,效率较高,说明其预测效果较好。

At present,big prediction errors of small sample data may affect the data application effect directly.In order to improve the prediction accuracy,this paper presented a method of designing a small sample data prediction model based on multivariate statistical analysis.At first,we put the small sample data into SPSS software,and used the bootstrap method to analyze the empirical distribution of small sample data.Based on the empirical distribution characteristics,we used the Fisher face algorithm with learning ability to pre-classify the data of small sample,and then constructed a test sample category label to extract the feature of small sample data.Through the principal components of data features based on multivariate statistical analysis,we determined the regression function of the model.Finally,we built a data prediction model in combination with a support vector machine.Thus,an accurate prediction was achieved.Experimental results show that the proposed method has lower prediction error and higher efficiency in predicting small sample data,indicating that the prediction effect is better.