dbs classification

{
 "metadata": {
  "name": "analysis_dbs"
 },
 "nbformat": 3,
 "nbformat_minor": 0,
 "worksheets": [
  {
   "cells": [
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "%load_ext autoreload\n",
      "%autoreload 2\n",
      "import hddm\n",
      "print hddm.__version__\n",
      "import kabuki\n",
      "print kabuki.__version__\n",
      "import pandas as pd\n",
      "import matplotlib.pyplot as plt\n",
      "from scipy import stats\n",
      "\n",
      "from sklearn.grid_search import GridSearchCV\n",
      "from sklearn.metrics import classification_report\n",
      "from sklearn.metrics import precision_score\n",
      "from sklearn.metrics import recall_score\n",
      "from sklearn.svm import SVC\n",
      "from sklearn.cross_validation import LeaveOneOut\n",
      "from sklearn import preprocessing"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "0.5dev\n",
        "0.5dev\n"
       ]
      }
     ],
     "prompt_number": 444
    },
    {
     "cell_type": "raw",
     "metadata": {},
     "source": [
      "Load and process data"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "data = hddm.load_csv('../data/PD_PS.csv')\n",
      "data['dbs_state'] = 'on'\n",
      "data['dbs_state'][data.dbs == 0] = 'off'"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 257
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "data_collapsed = data.copy()\n",
      "data_collapsed['subj_idx'] = np.int16(data_collapsed['subj_idx'])\n",
      "data_collapsed['subj_idx'].ix[data_collapsed.dbs_state == 'off'] += 14"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 285
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "Fit model"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "m = hddm.models.HDDMGamma(data_collapsed, depends_on={'v': 'conf', 'a': 'conf'}, plot_subjs=True)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 363
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "m.sample(5000, burn=2000)"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        " \r",
        "[****************100%******************]  5000 of 5000 complete"
       ]
      },
      {
       "output_type": "pyout",
       "prompt_number": 364,
       "text": [
        "<pymc.MCMC.MCMC at 0x10f2bf0c>"
       ]
      },
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "\n"
       ]
      }
     ],
     "prompt_number": 364
    },
    {
     "cell_type": "raw",
     "metadata": {},
     "source": [
      "Helper functions"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "def get_subj_nodes(m):\n",
      "    m.gen_stats()\n",
      "    nodes = m.nodes_db.ix[(m.nodes_db.stochastic == True) & (m.nodes_db.subj == True)]\n",
      "    return nodes\n",
      "\n",
      "def convert_to_features(m):\n",
      "    dfs = []\n",
      "    for subj, data in get_subj_nodes(m)[['mean', 'subj_idx', 'knode_name']].groupby(['subj_idx']):\n",
      "        reindex = [index.split('.')[0] for index in data.index]\n",
      "        data.index = reindex\n",
      "        dfs.append(pd.DataFrame(data['mean'], columns=[subj]).T)\n",
      "\n",
      "    f_vec = pd.concat(dfs)\n",
      "    return f_vec"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 384
    },
    {
     "cell_type": "markdown",
     "metadata": {},
     "source": [
      "Create feature vectores, scale etc"
     ]
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "X = convert_to_features(m).drop(['v_subj(HC)', 'v_subj(LC)'], axis=1)\n",
      "\n",
      "X_scaled = preprocessing.scale(X)\n",
      "y = np.concatenate([np.ones(14), np.zeros(14)])"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [],
     "prompt_number": 439
    },
    {
     "cell_type": "code",
     "collapsed": false,
     "input": [
      "param_grid = [\n",
      "  {'C': [.01, .1, 1, 10, 100, 1000], 'kernel': ['linear']},\n",
      "  {'C': [.01, .1, 1, 10, 100, 1000], 'gamma': [10, 1, 0.1, 0.01, 0.001, 0.0001], 'kernel': ['rbf']},\n",
      " ]\n",
      "\n",
      "clf = GridSearchCV(SVC(), param_grid, score_func=metrics.zero_one_score)\n",
      "clf.fit(X_scaled, y, cv=5)\n",
      "print 'Best model:', clf.best_estimator\n",
      "print 'Best cv score:', clf.best_score"
     ],
     "language": "python",
     "metadata": {},
     "outputs": [
      {
       "output_type": "stream",
       "stream": "stdout",
       "text": [
        "Best model: SVC(C=10, cache_size=200, coef0=0.0, degree=3, gamma=1, kernel=rbf,\n",
        "  probability=False, scale_C=False, shrinking=True, tol=0.001)\n",
        "Best cv score: 0.72\n"
       ]
      }
     ],
     "prompt_number": 443
    }
   ],
   "metadata": {}
  }
 ]
}