Conditional canonical correlation estimation based on covariates with random forests

Covariance regression with random forests

Capturing the conditional covariances or correlations among the elements of a multivariate response vector based on covariates is important to various fields including neuroscience, epidemiology and biomedicine. We propose a new method called Covariance Regression with Random Forests (CovRegRF) to estimate the covariance matrix of a multivariate response given a set of covariates, using a random forest framework. Random forest trees are built with a splitting rule specially designed to maximize the difference between the sample covariance matrix estimates of the child nodes. We also propose a significance test for the partial effect of a subset of covariates. We evaluate the performance of the proposed method and significance test through a simulation study which shows that the proposed method provides accurate covariance matrix estimates and that the Type-1 error is well controlled. An application of the proposed method to thyroid disease data is also presented. CovRegRF is implemented in a freely available R package on CRAN.

Multivariate associations between cognition and neighborhood geospatial characteristics in schizophrenia

Cognitive impairment contributes to functional impairment in schizophrenia. Yet, little is known about how environmental characteristics are related to cognition in schizophrenia. By examining how cognition and the environment are intertwined, it may be possible to identify modifiable risk and protective factors that can improve cognitive outcomes in schizophrenia. We aimed to identify multivariate associations between cognition and three geospatial characteristics (built-space density, habitable green spaces, and public spaces for social interaction) within one's immediate neighborhood among individuals with schizophrenia. We recruited participants with schizophrenia from three sites - an urban metropolitan and two towns in southern India. We administered standard cognitive assessments and performed a principal axis factoring to identify episodic memory, cognitive control, and social inference-making factors for use in further analyses. We estimated geospatial characteristics of an individual's neighborhood, i.e., up to 1 km² around the residence, by sourcing data from Google Earth. We performed unconditional and conditional (to examine the effect of clinical covariates) canonical correlation analyses to understand the multivariate relationship between cognition and geospatial characteristics. We analyzed data from 208 participants; the first canonical cognitive variate (higher social inference-making and poorer cognitive control) shared 24% of the variance (r = 0.49; P < 0.001) with the first geospatial variate (lower built density and poorer access to public spaces). Years of education, age at onset, and place of residence significantly modulated this relationship. We observe differential associations of the built environment with social and non-social cognition in schizophrenia, and highlight the clinical and demographic characteristics that shape these associations.

Machine learning due diligence evaluation to increase NPLs profitability transactions on secondary market

In this paper, we contribute to the topic of the non-performing loans (NPLs) business profitability on the secondary market by developing machine learning-based due diligence. In particular, a loan became non-performing when the borrower is unlikely to pay, and we use the ability of the ML algorithms to model complex relationships between predictors and outcome variables, we set up an ad hoc dependent random forest regressor algorithm for projecting the recovery rate of a portfolio of the secured NPLs. Indeed the profitability of the transactions under consideration depends on forecast models of the amount of net repayments expected from receivables and related collection times. Finally, the evaluation approach we provide helps to reduce the ”lemon discount” by pricing the risky component of informational asymmetry between better-informed banks and potential investors in particular for higher quality, collateralised NPLs.