Skip to main content
Fig. 4 | BMC Bioinformatics

Fig. 4

From: Standard machine learning approaches outperform deep representation learning on phenotype prediction from transcriptomics data

Fig. 4

Performance by supervised model. The classifier tasks are shown in the same format as Fig. 2. In this figure, the model results are divided into groups based on which supervised model is used, kNN (k-nearest neighbors), LR (logistic regression), or RF (random forest). The three horizontal lines for each task show the median result for each of these supervised models in this order (shown in the upper left of the first column plot). For each task, the supervised model with the best median result is shown at the bottom of the plot. While the median over the RF results is most frequently best (for thirteen tasks, compared to nine for logistic regression and two for k-nearest neighbors), the best performing logistic regression models are more consistently high performing among models

Back to article page