visualize_visit_embeddings: Visualize Embeddings of Visits

In adamgdobrakowski/memr: Multisource Embeddings for Medical Records

Description

Visualize given embeddings of visits of specified specialties of doctors. The visualization is generated on the 2D plane by t-SNE algorithm or by PCA (two main components) and plotted with ggplot. If t-SNE cannot be generated (because for example perplexity is too large for the number of samples), there are plotted two main components from PCA. Points representing the visits can be colored by doctor ID, the first letter of ICD-10 code or by cluster number. The plot can be optionally saved to the given PDF file.

Usage

visualize_visit_embeddings(visits_vectors, visit_table, method = "tsne",
  color_by, spec = NULL, clusters, n_doctors = 5L, n_codes = 9L,
  save = FALSE, path_to_save)

Arguments

visit_table	A data frame with columns: visit_id icd10 (Optional, if color_by == "icd10"): ICD-10 code of the visit doctor_id (Optional, if color_by == "doctor"): doctor's ID specialties Specialties of doctors, separated by ", "
method	One of "tsne" (default) or "pca" - a method of generating the plot
color_by	A string, one of {"doctor", "icd10", "cluster"}
spec	A specialty to be plotted (optional, required if color_by != "clusters")
clusters	An output of the function cluster_visits (optional, required if color_by == "clusters")
n_doctors	A number of doctors to be colored, the rest will be grey (optional, required if color_by == "doctor"), default: 5
n_codes	A number of codes to be colored, the rest will be grey (optional, required if color_by == "icd10"), default: 9
save	A logical indicating if the plot should be saved to the file
path_to_save	An optional string of the path to the target PDF file
visit_vectors	A matrix of embeddings of visits

Value

A generated plot of embeddings.

Examples

inter_term_vectors <- embed_terms(interviews, embedding_size = 10L,
                                  term_count_min = 1L)
exam_term_vectors <- embed_terms(examinations, embedding_size = 10L,
                                 term_count_min = 1L)
visits_vectors <- embed_list_visits(interviews,
                                    examinations, inter_term_vectors,
                                    exam_term_vectors)

visualize_visit_embeddings(visits_vectors, visits, "tsne",
                           color_by = "doctor", spec = "internist", n_doctors = 10)
visualize_visit_embeddings(visits_vectors, visits, "pca",
                           color_by = "doctor", spec = "internist", n_doctors = 9)
visualize_visit_embeddings(visits_vectors, visits,
                           color_by = "icd10", spec = "internist", n_codes = 4)

clusters <- cluster_visits(visits_vectors, visits,
                           spec = "internist", cluster_number = 2)
visualize_visit_embeddings(visits_vectors, visits,
                           color_by = "cluster",
                           clusters = clusters)

adamgdobrakowski/memr documentation built on Sept. 4, 2021, 3:45 a.m.