Statistical learning methods in data science and their empirical research
Tian Qi
Shopify (USA) Inc., New York
DOI: https://doi.org/10.59429/esta.v12i2.10578
Keywords: Data science; Statistical learning methods; Regression analysis; Clustering models; Classification models
Abstract
Statistical learning methods serve as critical tools in data science, playing a key role in the analysis and modeling of complex data. This study focuses on three statistical learning methods—regression, clustering, and classification—conducting empirical research using the UCI “Online Retail” dataset. Linear regression models were employed to evaluate the impact of key features on the performance of continuous variable prediction. Results show that the inclusion of “purchase frequency” improved the model’s goodness of fit (R2) to 0.87 and reduced the mean squared error (MSE) to 12.3. For clustering tasks, the K-Means algorithm was applied, and when K=3, the silhouette coefficient reached a maximum of 0.71, effectively distinguishing different consumption behavior patterns. In the classification task, the random forest model achieved an accuracy of 93.5% and an F1-score of 0.90, demonstrating its robustness in high-dimensional data scenarios. The findings highlight the significant advantages of different statistical learning methods in their respective tasks while pointing out areas for improvement, such as feature dependency, computational efficiency, and handling of outliers.
References
[1]Asatryan N M , Shmyr S I , Timofeev I B , et al.Development, study, and comparison of models of crossimmunity to the influenza virus using statistical methods and machine learning.[J].Voprosy virusologii, 2024, 69(4):349-362.
[2]Abdi F M , BabaAli B , Momeni S .An unsupervised statistical representation learning method for human activity recognition[J].Signal, Image and Video Processing, 2024, 18(10):7041-7052.
[3]Anne-Laure B , N M W , Sabine H , et al.Statistical learning approaches in the genetic epidemiology of complex diseases.[J].Human genetics, 2020, 139(1):73-84.
[4]Karim E M , Petkau J , Gustafson P , et al.On the application of statistical learning approaches to construct inverse probability weights in marginal structural Cox models: Hedging against weight-model misspecification[J].Communications in Statistics - Simulation and Computation, 2017, 46(10):7668-7697.
[5]Ishak B A , Feki A .A discriminating study between three categories of banks based on statistical learning approaches[J].Intelligent Data Analysis, 2016, 20(5):1199-1221.