On Combining Clusterwise Linear Regression and K-Means with Automatic Weighting of the Explanatory Variables

Silva, Ricardo A. M. da; Carvalho, Francisco de A. T. de

doi:10.1007/978-3-319-68612-7_46

Cited by 4 publications

(3 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We compare the performance of DoGR to existing stateof-the-art methods for disaggregated regression: the three variants of CLR, WCLR (da Silva and de Carvalho 2017), FWCLR (da Silva and de Carvalho 2018), and GMR (Sung 2004). We use CART as a method which uses decision tree for regression.…”

Section: Quantitative Resultsmentioning

confidence: 99%

See 1 more Smart Citation

DoGR: Disaggregated Gaussian Regression for Reproducible Analysis of Heterogeneous Data

Alipourfard,

Burghardt,

Lerman

2021

Preprint

View full text Add to dashboard Cite

Quantitative analysis of large-scale data is often complicated by the presence of diverse subgroups, which reduce the accuracy of inferences they make on held-out data. To address the challenge of heterogeneous data analysis, we introduce DoGR, a method that discovers latent confounders by simultaneously partitioning the data into overlapping clusters (disaggregation) and modeling the behavior within them (regression). When applied to real-world data, our method discovers meaningful clusters and their characteristic behaviors, thus giving insight into group differences and their impact on the outcome of interest. By accounting for latent confounders, our framework facilitates exploratory analysis of noisy, heterogeneous data and can be used to learn predictive models that better generalize to new data. We provide the code to enable others to use DoGR within their data analytic workflows 1 .

show abstract

Section: Quantitative Resultsmentioning

confidence: 99%

“…The method is slow, since it moves one data point at each iteration. Two other methods, WCLR (da Silva and de Carvalho 2017) and FWCLR (da Silva and de Carvalho 2018), improve on CLR by using k-means as their clustering method. These methods were shown to outperform CLR and other methods, such as K-plane (Manwani and Sastry 2015).…”

Section: Introductionmentioning

confidence: 99%

DoGR: Disaggregated Gaussian Regression for Reproducible Analysis of Heterogeneous Data

Alipourfard,

Burghardt,

Lerman

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…Controlling for these latent groups improves the utility of linear models. Methods like cluster-wise linear regression [36][37][38] split data into groups using unsupervised methods, like k-means, and then fit linear models within each cluster. However, such methods can be slow on large data.…”

Section: Identifying the Latent Componentsmentioning

confidence: 99%

Socioeconomic Correlates of Anti-Science Attitudes in the US

Hu¹,

Rao²,

Kejriwal³

et al. 2021

Future Internet

View full text Add to dashboard Cite

Successful responses to societal challenges require sustained behavioral change. However, as responses to the COVID-19 pandemic in the US showed, political partisanship and mistrust of science can reduce public willingness to adopt recommended behaviors such as wearing a mask or receiving a vaccination. To better understand this phenomenon, we explored attitudes toward science using social media posts (tweets) that were linked to counties in the US through their locations. The data allowed us to study how attitudes towards science relate to the socioeconomic characteristics of communities in places from which people tweet. Our analysis revealed three types of communities with distinct behaviors: those in large metro centers, smaller urban places, and rural areas. While partisanship and race are strongly associated with the share of anti-science users across all communities, income was negatively and positively associated with anti-science attitudes in suburban and rural areas, respectively. We observed that emotions in tweets, specifically negative high arousal emotions, are expressed among suburban and rural communities by many anti-science users, but not in communities in large urban places. These trends were not apparent when pooled across all counties. In addition, we found that anti-science attitudes expressed five years earlier were significantly associated with lower COVID-19 vaccination rates. Our analysis demonstrates the feasibility of using spatially resolved social media data to monitor public attitudes on issues of social importance.

show abstract

On Combining Fuzzy C-Regression Models and Fuzzy C-Means with Automated Weighting of the Explanatory Variables

Silva

Carvalho

2018

2018 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE)

View full text Add to dashboard Cite

On Combining Clusterwise Linear Regression and K-Means with Automatic Weighting of the Explanatory Variables

Cited by 4 publications

References 8 publications

DoGR: Disaggregated Gaussian Regression for Reproducible Analysis of Heterogeneous Data

DoGR: Disaggregated Gaussian Regression for Reproducible Analysis of Heterogeneous Data

Socioeconomic Correlates of Anti-Science Attitudes in the US

On Combining Fuzzy C-Regression Models and Fuzzy C-Means with Automated Weighting of the Explanatory Variables

Contact Info

Product

Resources

About