Performance comparison of linear and non-linear feature selection methods for the analysis of large survey datasets

摘要: Large survey databases for aging-related analysis are often examined to discover key factors that affect a dependent variable of interest. Typically, this is performed with methods assuming linear dependencies between variables. Such assumptions however do not hold in many cases, wherein data linked by way non-linear dependencies. This turn requires applications analytic methods, which more accurate identifying potentially Here, we objectively compared the feature selection performance several frequently-used and three context large data. These were assessed using both synthetic real-world datasets, relationships features variables known advance. In contrast found offered better overall than all usage conditions. Moreover, was stable, being unaffected inclusion or exclusion from datasets. properties make preferable tool hypothesis-driven exploratory analyses