mixtcomp-example.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# RMixtComp basic example"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data\n",
    "\n",
    "Load the CSV data file as dataframe."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "data <- read.table(\"mixtcomp-example.csv\", sep = \";\", header = TRUE)\n",
    "head(data)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Clustering with RMixtComp"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Launch the RMixtComp package."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "library(RMixtComp)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Define the distribution used for each variable."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model <- list(Age = \"Gaussian\", Wt = \"Gaussian\", PF = \"Multinomial\", \n",
    "              HX = \"Multinomial\", SBP = \"Gaussian\", DBP = \"Gaussian\", \n",
    "              EKG = \"Multinomial\", HG = \"Gaussian\", SZ = \"Gaussian\", \n",
    "              SG = \"Gaussian\", AP = \"Gaussian\", BM = \"Multinomial\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Define the SEM algorithm's parameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "algo <- list(nbBurnInIter = 50,\n",
    "             nbIter = 100,\n",
    "             nbGibbsBurnInIter = 50,\n",
    "             nbGibbsIter = 100,\n",
    "             nInitPerClass = floor(nrow(data)/2),\n",
    "             nSemTry = 5,\n",
    "             confidenceLevel = 0.95,\n",
    "             ratioStableCriterion = 0.99,\n",
    "             nStableCriterion = 10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Choose the desired number of classes and the number of runs for each given number of classes."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "nClass <- 1:8\n",
    "nRun <- 3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "res <- mixtCompLearn(data, model, algo, nClass = nClass, criterion = \"ICL\", nRun = nRun, nCore = 1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Output's Analysis"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Criterion"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Draw the criterion value (BIC and ICL) for each model that was built. The higher the value (close to 0) the better the model. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plotCrit(res, pkg = \"plotly\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Choose the number of classes to study in the following."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "K <- 3\n",
    "resK <- extractMixtCompObject(res, K)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Variables"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Draw the discriminating level of each variable. A high value (close to one) means that the variable is highly discriminating. A low value (close to zero) means that the variable is poorly discriminating."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plotDiscrimVar(resK, pkg = \"plotly\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Draw the similarity between every pair of variable. A high value (close to one) means that the two variables provide the same information for the clustering task (i.e. similar partitions). A low value (close to zero) means that the two variables provide some different information for the clustering task (i.e. different partitions)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "heatmapVar(resK, pkg = \"plotly\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Select a variable to draw its distribution."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "variable <- \"SG\"\n",
    "plotDataBoxplot(resK, variable, grl = TRUE, pkg = \"plotly\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Classes"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Draw the proportion of individuals in each class."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plotProportion(resK, pkg = \"plotly\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Draw the similarity level between each pair of classes. A high value (close to one) means that the 2 classes are strongly different (i.e. low overlapping). A low value (close to zero) means that the 2 classes are similar for the clustering task (i.e. high overlapping)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "heatmapClass(resK, pkg = \"plotly\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Draw the discriminating level of each variable for the selected class."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "class <- 2\n",
    "plotDiscrimVar(resK, class = class, pkg = \"plotly\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Select a variable to draw its distribution for the selected class."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "variable <- \"SG\"\n",
    "plotDataBoxplot(resK, variable, class = class, grl = TRUE, pkg = \"plotly\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Probabilities"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Draw the probability of assignment to a class for each individual. Individuals have been reordered in decreasing assignment probability. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "heatmapTikSorted(resK, pkg = \"plotly\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Advanced"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Visualize in a *Gaussian-like way*, and onto R2, results of Gaussian or non-Gaussian based clustering."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "library(ClusVis)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "logTik <- getTik(resK, log = TRUE)\n",
    "prop <- getProportion(resK)\n",
    "resVisu <- clusvis(logTik, prop)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Component Interpretation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plotDensityClusVisu(resVisu, add.obs = FALSE)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Observation Scatter-plot "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plotDensityClusVisu(resVisu, add.obs = TRUE)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "R",
   "language": "R",
   "name": "ir"
  },
  "language_info": {
   "codemirror_mode": "r",
   "file_extension": ".r",
   "mimetype": "text/x-r-source",
   "name": "R",
   "pygments_lexer": "r",
   "version": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}