RNA_seq_DNN_hyperparamter.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "RNA_seq_DNN_hyperparamter.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "authorship_tag": "ABX9TyPJCVEkWvzQe8m8EY6uRgf3",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/Aayushktyagi/DeepLearning_Resources/blob/master/RNA_seq_DNN_hyperparamter.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "R9blyEWvQQAR",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "5a382e76-5f96-48a0-cca7-17b1ba67f6be"
      },
      "source": [
        "from google.colab import drive\n",
        "drive.mount('/content/drive')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount(\"/content/drive\", force_remount=True).\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lR4kGLXOT1je"
      },
      "source": [
        "# Load the TensorBoard notebook extension\n",
        "%load_ext tensorboard\n",
        "\n",
        "# Clear any logs from previous runs\n",
        "!rm -rf ./logs/ "
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "l13q3jKRQ7FG"
      },
      "source": [
        "import time\n",
        "import numpy as np\n",
        "import pandas as pd \n",
        "import tensorflow as tf \n",
        "import sklearn \n",
        "import tensorflow.keras.backend as K\n",
        "from tensorflow import keras\n",
        "from tensorflow import feature_column\n",
        "from tensorflow.keras import layers\n",
        "from tensorflow.keras.callbacks import EarlyStopping, ModelCheckpoint, ReduceLROnPlateau\n",
        "from sklearn.preprocessing import LabelEncoder\n",
        "from sklearn import preprocessing\n",
        "from sklearn.decomposition import PCA\n",
        "from sklearn.manifold import TSNE\n",
        "import seaborn as sns\n",
        "from scipy import stats\n",
        "from sklearn.metrics import confusion_matrix\n",
        "from sklearn.metrics import accuracy_score\n",
        "from sklearn.metrics import classification_report\n",
        "import matplotlib.pyplot as plt\n",
        "import matplotlib.patheffects as PathEffects\n",
        "from sklearn.model_selection import train_test_split\n",
        "from keras.utils import np_utils\n",
        "from tensorboard.plugins.hparams import api as hp"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 342
        },
        "id": "KKCy8pBKQ-B9",
        "outputId": "2f1f0493-414a-47ae-90c9-871dc5e39270"
      },
      "source": [
        "filepath = '/content/drive/My Drive/ColabNotebooks/Cancer/RNA_seq_labeled_primary.csv'\n",
        "raw_data = pd.read_csv(filepath, sep = ',')\n",
        "raw_data.head()\n",
        "dataframe = raw_data.copy()\n",
        "# data = data.drop(data.index[0])\n",
        "dataframe.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>sampleID</th>\n",
              "      <th>ARHGEF10L</th>\n",
              "      <th>HIF3A</th>\n",
              "      <th>RNF17</th>\n",
              "      <th>RNF10</th>\n",
              "      <th>RNF11</th>\n",
              "      <th>RNF13</th>\n",
              "      <th>GTF2IP1</th>\n",
              "      <th>REM1</th>\n",
              "      <th>MTVR2</th>\n",
              "      <th>RTN4RL2</th>\n",
              "      <th>C16orf13</th>\n",
              "      <th>C16orf11</th>\n",
              "      <th>FGFR1OP2</th>\n",
              "      <th>TSKS</th>\n",
              "      <th>ATRX</th>\n",
              "      <th>PMM2</th>\n",
              "      <th>LOC100272146</th>\n",
              "      <th>ASS1</th>\n",
              "      <th>NCBP1</th>\n",
              "      <th>ZNF709</th>\n",
              "      <th>ZNF708</th>\n",
              "      <th>RBM14</th>\n",
              "      <th>NCBP2</th>\n",
              "      <th>DISC1</th>\n",
              "      <th>CAMK1</th>\n",
              "      <th>RPL37</th>\n",
              "      <th>SPR</th>\n",
              "      <th>ZNF700</th>\n",
              "      <th>ZNF707</th>\n",
              "      <th>CAMK4</th>\n",
              "      <th>ZNF704</th>\n",
              "      <th>LOC339240</th>\n",
              "      <th>GOLGA6B</th>\n",
              "      <th>RNF115</th>\n",
              "      <th>RNF112</th>\n",
              "      <th>ZC3H14</th>\n",
              "      <th>SPN</th>\n",
              "      <th>HMGCLL1</th>\n",
              "      <th>NACAP1</th>\n",
              "      <th>...</th>\n",
              "      <th>ACAP2</th>\n",
              "      <th>ACAP3</th>\n",
              "      <th>LUZP4</th>\n",
              "      <th>OR10J5</th>\n",
              "      <th>LUZP6</th>\n",
              "      <th>C1orf198</th>\n",
              "      <th>CFL2</th>\n",
              "      <th>CFL1</th>\n",
              "      <th>OR10J3</th>\n",
              "      <th>C8orf83</th>\n",
              "      <th>C1orf192</th>\n",
              "      <th>PLAA</th>\n",
              "      <th>C8orf80</th>\n",
              "      <th>C8orf86</th>\n",
              "      <th>C8orf85</th>\n",
              "      <th>C8orf84</th>\n",
              "      <th>NFIC</th>\n",
              "      <th>NFIB</th>\n",
              "      <th>H1FOO</th>\n",
              "      <th>NFIX</th>\n",
              "      <th>PLEKHG6</th>\n",
              "      <th>PLEKHG7</th>\n",
              "      <th>PLEKHG4</th>\n",
              "      <th>PLEKHG5</th>\n",
              "      <th>SLC7A14</th>\n",
              "      <th>SELE</th>\n",
              "      <th>SLC7A10</th>\n",
              "      <th>PLA2G2C</th>\n",
              "      <th>METTL11A</th>\n",
              "      <th>TULP2</th>\n",
              "      <th>NPY5R</th>\n",
              "      <th>GNGT2</th>\n",
              "      <th>GNGT1</th>\n",
              "      <th>TULP3</th>\n",
              "      <th>PTRF</th>\n",
              "      <th>BCL6B</th>\n",
              "      <th>GSTK1</th>\n",
              "      <th>SELP</th>\n",
              "      <th>SELS</th>\n",
              "      <th>label</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>TCGA-CA-5256-01</td>\n",
              "      <td>10.1616</td>\n",
              "      <td>3.7172</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>10.4857</td>\n",
              "      <td>10.5157</td>\n",
              "      <td>11.1805</td>\n",
              "      <td>11.3755</td>\n",
              "      <td>2.7680</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>2.2321</td>\n",
              "      <td>10.4496</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>8.8640</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>9.3775</td>\n",
              "      <td>11.2484</td>\n",
              "      <td>6.1007</td>\n",
              "      <td>13.4524</td>\n",
              "      <td>10.7642</td>\n",
              "      <td>6.7441</td>\n",
              "      <td>7.7266</td>\n",
              "      <td>9.9810</td>\n",
              "      <td>11.9751</td>\n",
              "      <td>4.8262</td>\n",
              "      <td>8.6103</td>\n",
              "      <td>14.4843</td>\n",
              "      <td>10.9548</td>\n",
              "      <td>8.4412</td>\n",
              "      <td>6.9918</td>\n",
              "      <td>3.0698</td>\n",
              "      <td>6.5216</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>8.1044</td>\n",
              "      <td>1.3702</td>\n",
              "      <td>9.5912</td>\n",
              "      <td>5.5207</td>\n",
              "      <td>2.0600</td>\n",
              "      <td>7.5548</td>\n",
              "      <td>...</td>\n",
              "      <td>9.8673</td>\n",
              "      <td>8.4951</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>11.8241</td>\n",
              "      <td>9.1441</td>\n",
              "      <td>7.6565</td>\n",
              "      <td>14.5395</td>\n",
              "      <td>0.0</td>\n",
              "      <td>10.4282</td>\n",
              "      <td>1.8646</td>\n",
              "      <td>9.9139</td>\n",
              "      <td>2.7680</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>1.0403</td>\n",
              "      <td>6.6637</td>\n",
              "      <td>7.5262</td>\n",
              "      <td>8.4565</td>\n",
              "      <td>0.0</td>\n",
              "      <td>9.3462</td>\n",
              "      <td>10.5922</td>\n",
              "      <td>3.7740</td>\n",
              "      <td>8.6320</td>\n",
              "      <td>8.5671</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>4.1624</td>\n",
              "      <td>1.3702</td>\n",
              "      <td>0.0</td>\n",
              "      <td>10.2633</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>4.0747</td>\n",
              "      <td>0.6119</td>\n",
              "      <td>8.0334</td>\n",
              "      <td>9.4792</td>\n",
              "      <td>5.7645</td>\n",
              "      <td>12.5517</td>\n",
              "      <td>3.9322</td>\n",
              "      <td>10.8937</td>\n",
              "      <td>Stage IIA</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>TCGA-AZ-6599-01</td>\n",
              "      <td>11.1212</td>\n",
              "      <td>2.3437</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>11.6878</td>\n",
              "      <td>10.3830</td>\n",
              "      <td>9.9166</td>\n",
              "      <td>11.9240</td>\n",
              "      <td>2.4817</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>3.4169</td>\n",
              "      <td>9.7643</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>8.8247</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>10.0715</td>\n",
              "      <td>10.5191</td>\n",
              "      <td>5.5343</td>\n",
              "      <td>12.1302</td>\n",
              "      <td>10.6109</td>\n",
              "      <td>6.2459</td>\n",
              "      <td>6.7480</td>\n",
              "      <td>10.4944</td>\n",
              "      <td>10.8831</td>\n",
              "      <td>6.5712</td>\n",
              "      <td>6.9435</td>\n",
              "      <td>13.9471</td>\n",
              "      <td>10.3406</td>\n",
              "      <td>8.1277</td>\n",
              "      <td>7.2890</td>\n",
              "      <td>4.9331</td>\n",
              "      <td>7.9725</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>8.0904</td>\n",
              "      <td>2.4817</td>\n",
              "      <td>10.4349</td>\n",
              "      <td>5.6704</td>\n",
              "      <td>0.5941</td>\n",
              "      <td>7.4532</td>\n",
              "      <td>...</td>\n",
              "      <td>9.7920</td>\n",
              "      <td>9.5212</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>11.1356</td>\n",
              "      <td>10.2243</td>\n",
              "      <td>6.9869</td>\n",
              "      <td>14.2384</td>\n",
              "      <td>0.0</td>\n",
              "      <td>10.2780</td>\n",
              "      <td>4.0699</td>\n",
              "      <td>9.2377</td>\n",
              "      <td>1.3383</td>\n",
              "      <td>0.5941</td>\n",
              "      <td>1.3383</td>\n",
              "      <td>10.4418</td>\n",
              "      <td>8.9842</td>\n",
              "      <td>9.3824</td>\n",
              "      <td>0.0</td>\n",
              "      <td>11.3403</td>\n",
              "      <td>9.8757</td>\n",
              "      <td>5.8970</td>\n",
              "      <td>10.6252</td>\n",
              "      <td>8.7413</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>2.7234</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>0.0</td>\n",
              "      <td>8.9257</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.5941</td>\n",
              "      <td>1.8268</td>\n",
              "      <td>1.3383</td>\n",
              "      <td>8.3262</td>\n",
              "      <td>8.9279</td>\n",
              "      <td>4.8842</td>\n",
              "      <td>11.6017</td>\n",
              "      <td>4.1549</td>\n",
              "      <td>10.5855</td>\n",
              "      <td>Stage I</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>TCGA-AA-3655-01</td>\n",
              "      <td>11.0245</td>\n",
              "      <td>2.0858</td>\n",
              "      <td>0.5495</td>\n",
              "      <td>11.5062</td>\n",
              "      <td>9.9334</td>\n",
              "      <td>9.6560</td>\n",
              "      <td>11.6354</td>\n",
              "      <td>3.7064</td>\n",
              "      <td>0.9465</td>\n",
              "      <td>3.3607</td>\n",
              "      <td>9.4322</td>\n",
              "      <td>0.5495</td>\n",
              "      <td>8.6209</td>\n",
              "      <td>0.9465</td>\n",
              "      <td>11.0433</td>\n",
              "      <td>10.9326</td>\n",
              "      <td>5.9695</td>\n",
              "      <td>12.5693</td>\n",
              "      <td>10.6616</td>\n",
              "      <td>6.0423</td>\n",
              "      <td>6.8519</td>\n",
              "      <td>10.6599</td>\n",
              "      <td>10.5108</td>\n",
              "      <td>7.4840</td>\n",
              "      <td>7.7608</td>\n",
              "      <td>13.1868</td>\n",
              "      <td>10.1857</td>\n",
              "      <td>8.1032</td>\n",
              "      <td>6.9740</td>\n",
              "      <td>2.3709</td>\n",
              "      <td>7.4690</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.5495</td>\n",
              "      <td>7.9735</td>\n",
              "      <td>2.4947</td>\n",
              "      <td>10.1151</td>\n",
              "      <td>6.8577</td>\n",
              "      <td>1.7303</td>\n",
              "      <td>6.9253</td>\n",
              "      <td>...</td>\n",
              "      <td>10.2136</td>\n",
              "      <td>8.9868</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>12.3580</td>\n",
              "      <td>10.5249</td>\n",
              "      <td>7.3346</td>\n",
              "      <td>14.7722</td>\n",
              "      <td>0.0</td>\n",
              "      <td>9.3289</td>\n",
              "      <td>3.4853</td>\n",
              "      <td>9.7248</td>\n",
              "      <td>1.9190</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>1.5132</td>\n",
              "      <td>5.1793</td>\n",
              "      <td>9.0101</td>\n",
              "      <td>9.6241</td>\n",
              "      <td>0.0</td>\n",
              "      <td>11.0373</td>\n",
              "      <td>9.6741</td>\n",
              "      <td>2.9917</td>\n",
              "      <td>9.5785</td>\n",
              "      <td>9.7169</td>\n",
              "      <td>0.5495</td>\n",
              "      <td>5.7998</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>0.0</td>\n",
              "      <td>9.2596</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>3.9851</td>\n",
              "      <td>0.5495</td>\n",
              "      <td>8.9149</td>\n",
              "      <td>10.6716</td>\n",
              "      <td>7.0262</td>\n",
              "      <td>12.4103</td>\n",
              "      <td>6.0321</td>\n",
              "      <td>10.3551</td>\n",
              "      <td>Stage II</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>TCGA-A6-6137-01</td>\n",
              "      <td>11.0576</td>\n",
              "      <td>6.0759</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>11.7053</td>\n",
              "      <td>10.2443</td>\n",
              "      <td>10.2250</td>\n",
              "      <td>11.3220</td>\n",
              "      <td>2.8994</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>4.0206</td>\n",
              "      <td>9.1701</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>8.7443</td>\n",
              "      <td>0.5474</td>\n",
              "      <td>10.7429</td>\n",
              "      <td>10.8209</td>\n",
              "      <td>5.9307</td>\n",
              "      <td>12.7200</td>\n",
              "      <td>10.4519</td>\n",
              "      <td>6.0759</td>\n",
              "      <td>7.1520</td>\n",
              "      <td>10.5990</td>\n",
              "      <td>11.5216</td>\n",
              "      <td>6.2953</td>\n",
              "      <td>7.7983</td>\n",
              "      <td>13.1937</td>\n",
              "      <td>10.1062</td>\n",
              "      <td>8.2577</td>\n",
              "      <td>8.2023</td>\n",
              "      <td>4.1387</td>\n",
              "      <td>8.4432</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.9434</td>\n",
              "      <td>8.2984</td>\n",
              "      <td>3.5941</td>\n",
              "      <td>10.2082</td>\n",
              "      <td>8.0598</td>\n",
              "      <td>2.0809</td>\n",
              "      <td>6.9834</td>\n",
              "      <td>...</td>\n",
              "      <td>10.4387</td>\n",
              "      <td>9.2880</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>11.7345</td>\n",
              "      <td>9.7900</td>\n",
              "      <td>7.2861</td>\n",
              "      <td>14.5882</td>\n",
              "      <td>0.0</td>\n",
              "      <td>9.6269</td>\n",
              "      <td>4.2124</td>\n",
              "      <td>9.5920</td>\n",
              "      <td>4.2825</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>2.2303</td>\n",
              "      <td>4.6169</td>\n",
              "      <td>9.0023</td>\n",
              "      <td>9.7930</td>\n",
              "      <td>0.0</td>\n",
              "      <td>11.4779</td>\n",
              "      <td>10.8237</td>\n",
              "      <td>4.1004</td>\n",
              "      <td>10.4159</td>\n",
              "      <td>8.7035</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>5.4719</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>0.0</td>\n",
              "      <td>9.0675</td>\n",
              "      <td>0.0</td>\n",
              "      <td>4.2124</td>\n",
              "      <td>3.7507</td>\n",
              "      <td>1.7258</td>\n",
              "      <td>8.4940</td>\n",
              "      <td>10.5038</td>\n",
              "      <td>6.9352</td>\n",
              "      <td>11.2349</td>\n",
              "      <td>6.3616</td>\n",
              "      <td>10.2726</td>\n",
              "      <td>Stage IIIB</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>TCGA-CK-4952-01</td>\n",
              "      <td>10.5660</td>\n",
              "      <td>1.9506</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>11.8684</td>\n",
              "      <td>9.8537</td>\n",
              "      <td>9.9356</td>\n",
              "      <td>11.4306</td>\n",
              "      <td>2.6602</td>\n",
              "      <td>0.4950</td>\n",
              "      <td>4.6048</td>\n",
              "      <td>9.3126</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>9.1254</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>9.8556</td>\n",
              "      <td>10.9294</td>\n",
              "      <td>5.0941</td>\n",
              "      <td>12.0369</td>\n",
              "      <td>11.0042</td>\n",
              "      <td>4.5301</td>\n",
              "      <td>5.7496</td>\n",
              "      <td>11.0857</td>\n",
              "      <td>10.5637</td>\n",
              "      <td>6.4625</td>\n",
              "      <td>6.5073</td>\n",
              "      <td>13.5306</td>\n",
              "      <td>9.4557</td>\n",
              "      <td>7.8924</td>\n",
              "      <td>8.4951</td>\n",
              "      <td>1.1557</td>\n",
              "      <td>7.5645</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>8.1717</td>\n",
              "      <td>2.3486</td>\n",
              "      <td>10.4859</td>\n",
              "      <td>5.0049</td>\n",
              "      <td>1.7891</td>\n",
              "      <td>7.9097</td>\n",
              "      <td>...</td>\n",
              "      <td>9.5390</td>\n",
              "      <td>9.9513</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>11.3376</td>\n",
              "      <td>10.5269</td>\n",
              "      <td>7.6075</td>\n",
              "      <td>14.3301</td>\n",
              "      <td>0.0</td>\n",
              "      <td>9.3664</td>\n",
              "      <td>3.5912</td>\n",
              "      <td>9.8652</td>\n",
              "      <td>1.3991</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>0.4950</td>\n",
              "      <td>10.8802</td>\n",
              "      <td>8.4951</td>\n",
              "      <td>9.4498</td>\n",
              "      <td>0.0</td>\n",
              "      <td>10.7055</td>\n",
              "      <td>9.9930</td>\n",
              "      <td>3.4362</td>\n",
              "      <td>9.4035</td>\n",
              "      <td>8.6309</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>2.8359</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>0.0</td>\n",
              "      <td>9.4047</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>2.8359</td>\n",
              "      <td>0.0000</td>\n",
              "      <td>9.4633</td>\n",
              "      <td>10.4412</td>\n",
              "      <td>6.7306</td>\n",
              "      <td>10.5621</td>\n",
              "      <td>3.6394</td>\n",
              "      <td>9.9296</td>\n",
              "      <td>Stage IIIC</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>5 rows × 20532 columns</p>\n",
              "</div>"
            ],
            "text/plain": [
              "          sampleID  ARHGEF10L   HIF3A  ...    SELP     SELS       label\n",
              "0  TCGA-CA-5256-01    10.1616  3.7172  ...  3.9322  10.8937   Stage IIA\n",
              "1  TCGA-AZ-6599-01    11.1212  2.3437  ...  4.1549  10.5855     Stage I\n",
              "2  TCGA-AA-3655-01    11.0245  2.0858  ...  6.0321  10.3551    Stage II\n",
              "3  TCGA-A6-6137-01    11.0576  6.0759  ...  6.3616  10.2726  Stage IIIB\n",
              "4  TCGA-CK-4952-01    10.5660  1.9506  ...  3.6394   9.9296  Stage IIIC\n",
              "\n",
              "[5 rows x 20532 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 4
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "qFM7vEDlRAr5",
        "outputId": "b9ac7df4-bd05-40a8-9cb7-bd4ef90617d6"
      },
      "source": [
        "dataframe['label'] = dataframe['label'].map({'Stage I':1, 'Stage II':2, 'Stage III':3,\n",
        "                                                                  'Stage IV':4, 'Stage IA':1,'Stage IIA':2,\n",
        "                                                                  'Stage IIB':2, 'Stage IIC':2, 'Stage IIIA':3,\n",
        "                                                                  'Stage IIIB':3,'Stage IIIC':3, 'Stage IVA':4,'Stage IVB':4,'[Discrepancy]':0 }\n",
        "                                                                 )\n",
        "\n",
        "dataframe['label'].value_counts()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "2.0    110\n",
              "3.0     80\n",
              "1.0     45\n",
              "4.0     39\n",
              "0.0      3\n",
              "Name: label, dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 5
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "a_-9B46KRDWS",
        "outputId": "1ff7b5d0-e234-4ac4-a783-8b1696dcc95b"
      },
      "source": [
        "# drop genes where all values are zero\n",
        "dataframe = dataframe.loc[:, (dataframe != 0).any()]\n",
        "dataframe.shape"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(286, 20037)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 6
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "xgocCgT-RGOM",
        "outputId": "d963a2d1-45bb-4869-d1c5-de940b1bc1bf"
      },
      "source": [
        "# check the rows/columns there sum is zero\n",
        "df_check = dataframe.loc[(dataframe.sum(axis=1) != 0), (dataframe.sum(axis=0) != 0)]\n",
        "df_check.shape"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(286, 20037)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 7
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XaGGIEVMRIcd"
      },
      "source": [
        "# drop rows with NAN\n",
        "dataframe = dataframe[dataframe['label'].notna()]\n",
        "# drop label where 0\n",
        "dataframe = dataframe[dataframe['label'] != 0]"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YdtUwjM7RKg6"
      },
      "source": [
        "cat_label = dataframe.pop('label')\n",
        "# drop gene id\n",
        "dataframe = dataframe.drop(['sampleID'],axis = 1)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "EzSEeY4DRNMZ",
        "outputId": "786d93ff-c15a-4228-9447-3a97c6859146"
      },
      "source": [
        "# Dictionary for key value pair of labels \n",
        "num_classes = np.unique(cat_label)\n",
        "le = preprocessing.LabelEncoder()\n",
        "le.fit(cat_label)\n",
        "print(list(le.classes_))\n",
        "label = le.transform(cat_label)\n",
        "# print(\"label:{}\".format(label))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[1.0, 2.0, 3.0, 4.0]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "10ohbLDCRR7Q",
        "outputId": "d0c52fca-9b2a-4291-d9ed-8b3883c03577"
      },
      "source": [
        "print(dataframe.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(274, 20035)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 265
        },
        "id": "gTIpCcvaRU7c",
        "outputId": "623a39c5-3eed-4515-dac2-7549b0121d3f"
      },
      "source": [
        "# data visualization \n",
        "plt.hist(label)\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "u4THcHdeRVr6"
      },
      "source": [
        "def data_scatter(x, colors):\n",
        "    # choose a color palette with seaborn.\n",
        "    num_classes = len(np.unique(colors))\n",
        "    palette = np.array(sns.color_palette(\"hls\", num_classes))\n",
        "\n",
        "    # create a scatter plot.\n",
        "    f = plt.figure(figsize=(8, 8))\n",
        "    ax = plt.subplot(aspect='equal')\n",
        "    sc = ax.scatter(x[:,0], x[:,1], lw=0, s=40, c=palette[colors.astype(np.int)])\n",
        "    plt.xlim(-25, 25)\n",
        "    plt.ylim(-25, 25)\n",
        "    ax.axis('off')\n",
        "    ax.axis('tight')\n",
        "\n",
        "    # add the labels for each digit corresponding to the label\n",
        "    txts = []\n",
        "\n",
        "    for i in range(num_classes):\n",
        "\n",
        "        # Position of each label at median of data points.\n",
        "\n",
        "        xtext, ytext = np.median(x[colors == i, :], axis=0)\n",
        "        txt = ax.text(xtext, ytext, str(i), fontsize=24)\n",
        "        txt.set_path_effects([\n",
        "            PathEffects.Stroke(linewidth=5, foreground=\"w\"),\n",
        "            PathEffects.Normal()])\n",
        "        txts.append(txt)\n",
        "\n",
        "    return f, ax, sc, txts"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZA6XxmrKRYg8"
      },
      "source": [
        "# Data Visualization \n",
        "def dim_reduction(dataframe):\n",
        "    pca = PCA(n_components = 50)\n",
        "    pca_result_50 = pca.fit_transform(dataframe)\n",
        "    RS = 123\n",
        "    start_time = time.time()\n",
        "    data_pca_tsne = TSNE(random_state=RS).fit_transform(pca_result_50)\n",
        "    data_scatter(data_pca_tsne , label)\n",
        "    print(\"Time taken:{}\".format(time.time() - start_time))\n",
        "    return pca_result_50"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 483
        },
        "id": "I0b5x20eRbpY",
        "outputId": "4c458655-5e42-42a3-9baa-5f6a5a572e43"
      },
      "source": [
        "# Data visualization for un normalized data\n",
        "data_pca = dim_reduction(dataframe)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Time taken:2.0522775650024414\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 576x576 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "CoagwSUeReBy",
        "outputId": "9521a863-12ca-4028-f9e2-531468d9aca0"
      },
      "source": [
        "data_des = dataframe.describe()\n",
        "data_des_transpose = data_des.T['std']\n",
        "zero_columns = np.where(data_des_transpose.where(data_des_transpose ==0).notnull().tolist())\n",
        "#drop columns based on idexes\n",
        "dataframe = dataframe.drop(dataframe.columns[zero_columns],axis=1)\n",
        "dataframe.shape"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(274, 20025)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 16
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "25FcZr4mRgdt",
        "outputId": "eec30b77-a5d8-48d0-e2db-0dad219d2981"
      },
      "source": [
        "# Data normalization \n",
        "# # Sklearn minmax scaler \n",
        "\n",
        "def normal_dataframe(dataframe , norm_type  , normalize = False):\n",
        "    if normalize == True:\n",
        "        if norm_type =='min_max':\n",
        "            min_max_scaler = preprocessing.MinMaxScaler()\n",
        "            dataframe_scaled = min_max_scaler.fit_transform(dataframe)\n",
        "            return dataframe_scaled\n",
        "        \n",
        "        elif norm_type == 'Stand_scaler':\n",
        "            scaler = preprocessing.StandardScaler()\n",
        "            dataframe_scaled = scaler.fit_transform(dataframe)\n",
        "            return dataframe_scaled\n",
        "        \n",
        "        elif norm_type == 'zscore':\n",
        "            # data_stats = dataframe.describe()\n",
        "            # print(data_stats)\n",
        "            # dataframe_scaled = (dataframe - data_stats['mean'])/ dataframe['std']\n",
        "            dataframe_scaled = stats.zscore(dataframe)\n",
        "            return dataframe_scaled\n",
        "        \n",
        "        elif norm_type == 'MaxAbsScaler':\n",
        "            max_abs_scaler = preprocessing.MaxAbsScaler()\n",
        "            dataframe_scaled = max_abs_scaler.fit_transform(dataframe)\n",
        "            return dataframe_scaled\n",
        "            \n",
        "        elif norm_type == 'log_magnitude':\n",
        "            dataframe_scaled = dataframe.apply(lambda x : np.log(x+1)) \n",
        "            return dataframe_scaled.to_numpy()\n",
        "        \n",
        "        elif norm_type == 'log_normalize':\n",
        "            dataframe_scaled = dataframe.apply(lambda x : np.log(x+1))\n",
        "            nlise = preprocessing.Normalizer()\n",
        "            dataframe_normalized = nlise.fit_transform(dataframe_scaled)\n",
        "            return dataframe_normalized\n",
        "    else:\n",
        "        return dataframe.to_numpy()\n",
        "    \n",
        "dataframe_scaled = normal_dataframe(dataframe ,'log_normalize', True)\n",
        "print(dataframe_scaled.shape)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(274, 20025)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 483
        },
        "id": "3L5cXhM5Rqzq",
        "outputId": "e995b915-9416-44ef-944c-0f66314bad64"
      },
      "source": [
        "# Data visualiazation after normalization \n",
        "data_pca = dim_reduction(dataframe_scaled)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Time taken:1.8459928035736084\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 576x576 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "LSFI37-mSgwo",
        "outputId": "1a7599bb-bb53-46c4-8c49-130c82fb2abc"
      },
      "source": [
        "#Test train Split\n",
        "# dataframe = dataframe.drop(['gene_id'],axis = 1)\n",
        "train, test , Y_train , Y_test = train_test_split(dataframe_scaled, label ,test_size = 0.1)\n",
        "train , val , Y_train , Y_val = train_test_split(train ,Y_train , test_size = 0.1)\n",
        "print(\"trainig data:{}\".format(len(train)))\n",
        "print(\"Validation data:{}\".format(len(val)))\n",
        "print(\"Testing data:{}\".format(len(test)))\n",
        "# print(\"Data Sample:{}\".format(train[0]))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "trainig data:221\n",
            "Validation data:25\n",
            "Testing data:28\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UnBuPPA0SjtW"
      },
      "source": [
        "def label_encoder(label):\n",
        "  # encode class values as integers\n",
        "  encoder = LabelEncoder()\n",
        "  encoder.fit(label)\n",
        "  encoded_Y = encoder.transform(label)\n",
        "  # convert integers to dummy variables (i.e. one hot encoded)\n",
        "  dummy_y = np_utils.to_categorical(encoded_Y)\n",
        "  return dummy_y"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bM04QtB2SmWK"
      },
      "source": [
        "# creating label one hot encoding\n",
        "train_label_oh = label_encoder(Y_train)\n",
        "test_label_oh = label_encoder(Y_test)\n",
        "val_label_oh = label_encoder(Y_val)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3aP9LDQkVGKK"
      },
      "source": [
        "# HP_NUM_UNITS = hp.HParam('num_units', hp.Discrete([16, 32]))\n",
        "HP_LEARNING_RATE =  hp.HParam('learning_rate',hp.Discrete([0.001,0.01,0.1]))\n",
        "HP_DROPOUT = hp.HParam('dropout', hp.RealInterval(0.1, 0.75))\n",
        "HP_OPTIMIZER = hp.HParam('optimizer', hp.Discrete(['adam']))\n",
        "\n",
        "METRIC_ACCURACY = 'accuracy'\n",
        "\n",
        "with tf.summary.create_file_writer('logs/hparam_tuning').as_default():\n",
        "  hp.hparams_config(\n",
        "    hparams=[ HP_LEARNING_RATE,HP_DROPOUT, HP_OPTIMIZER],\n",
        "    metrics=[hp.Metric(METRIC_ACCURACY, display_name='Accuracy')],\n",
        "  )"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lH-UWwhISqEW"
      },
      "source": [
        "#Base model \n",
        "def build_model_dnn(hparams):\n",
        "  model = tf.keras.Sequential([\n",
        "      #layer1 - input layer\n",
        "      layers.Dense(10000 , input_shape= (20025,) , name = 'Input'),\n",
        "      # adding batchnormalization\n",
        "      layers.BatchNormalization(),\n",
        "      layers.Activation('relu'),\n",
        "      layers.Dropout(hparams[HP_DROPOUT]),\n",
        "      # layers 2 hidden layer\n",
        "      layers.Dense(1000, name = \"Hidden_Layer_2\"),\n",
        "      layers.BatchNormalization(),\n",
        "      layers.Activation('relu'),\n",
        "      layers.Dropout(0.5),\n",
        "      layers.Dense(1000,  name = 'Hidden_layer_3'),\n",
        "      layers.BatchNormalization(),\n",
        "      layers.Activation('relu'),\n",
        "      layers.Dropout(0.5),\n",
        "      layers.Dense(100 , name = 'Hidden_layer_4'),\n",
        "      layers.BatchNormalization(),\n",
        "      layers.Activation('relu'),\n",
        "      layers.Dense(4 , activation = 'softmax' , name = \"Output_layer\")\n",
        "  ])\n",
        "  optimizer_name = hparams[HP_OPTIMIZER]\n",
        "  learning_rate = hparams[HP_LEARNING_RATE]\n",
        "  if optimizer_name == \"adam\":\n",
        "    optimizer = tf.keras.optimizers.Adam(learning_rate=learning_rate)\n",
        "  elif optimizer_name == \"sgd\":\n",
        "    optimizer = tf.keras.optimizers.SGD(learning_rate=learning_rate)\n",
        "  else:\n",
        "    raise ValueError(\"unexpected optimizer name: %r\" % (optimizer_name,))\n",
        "\n",
        "  model.compile(loss = 'categorical_crossentropy',\n",
        "               optimizer = optimizer,\n",
        "               metrics = ['accuracy'])\n",
        "  callbacks = [\n",
        "    EarlyStopping(patience=10, verbose=1,monitor='val_loss'),\n",
        "    ReduceLROnPlateau(factor=0.1, patience=3, min_lr=0.00001, verbose=1,monitor='val_loss'),\n",
        "    # tf.keras.callbacks.TensorBoard(logdir),\n",
        "    # hp.KerasCallback(logdir, hparams)\n",
        "    ]\n",
        "\n",
        "    #Model training \n",
        "  history = model.fit(\n",
        "            train, train_label_oh,\n",
        "            validation_data=(val, val_label_oh),\n",
        "            epochs = 100,\n",
        "            batch_size = 32,\n",
        "            callbacks=callbacks,\n",
        "            verbose=1)\n",
        "  \n",
        "  _, accuracy = model.evaluate(test, test_label_oh)\n",
        "  del model\n",
        "\n",
        "  return accuracy\n",
        "# model = build_model_dnn(hparams)\n",
        "# model.summary()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Rg0GG69nY0Mk"
      },
      "source": [
        "def run(run_dir, hparams):\n",
        "  with tf.summary.create_file_writer(run_dir).as_default():\n",
        "    hp.hparams(hparams)  # record the values used in this trial\n",
        "    accuracy = build_model_dnn(hparams)\n",
        "    tf.summary.scalar(METRIC_ACCURACY, accuracy, step=1)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pZkB6nseYaad"
      },
      "source": [
        "session_num = 0\n",
        "\n",
        "for lr in HP_LEARNING_RATE.domain.values:\n",
        "  for dropout_rate in (HP_DROPOUT.domain.min_value, HP_DROPOUT.domain.max_value):\n",
        "    for optimizer in HP_OPTIMIZER.domain.values:\n",
        "      hparams = {\n",
        "          HP_LEARNING_RATE: lr,\n",
        "          HP_DROPOUT: dropout_rate,\n",
        "          HP_OPTIMIZER: optimizer,\n",
        "      }\n",
        "      run_name = \"run-%d\" % session_num\n",
        "      print('--- Starting trial: %s' % run_name)\n",
        "      print({h.name: hparams[h] for h in hparams})\n",
        "      run('logs/hparam_tuning/' + run_name, hparams)\n",
        "      session_num += 1"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "HtlAPDfJaz93"
      },
      "source": [
        "%tensorboard --logdir logs/hparam_tuning"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zmE29sROTVuE"
      },
      "source": [
        "def predict(test):\n",
        "    predictions = model.predict(test)\n",
        "    predict_index = predictions.argmax(axis=1)\n",
        "    # print(predict_index,np.shape(predict_index))\n",
        "    test_index = test_label_oh.argmax(axis=1)\n",
        "    # print(test_index,np.shape(test_index))\n",
        "    results = confusion_matrix(test_index,predict_index)\n",
        "\n",
        "\n",
        "    #Classification accuracy\n",
        "    accuracy = accuracy_score(test_index,predict_index)\n",
        "    print(\"Classification accuracy:{}\".format(accuracy))\n",
        "    print()\n",
        "    print(accuracy)\n",
        "\n",
        "    #Classification report \n",
        "    classify_report = classification_report(test_index,predict_index)\n",
        "    print(\"Classification report\")\n",
        "    print(classify_report)\n",
        "\n",
        "# predict(test)"
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}