main_code.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "9eee623d-b23d-44de-887c-6b23e37f9a35",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.metrics import accuracy_score\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from sklearn.preprocessing import MinMaxScaler\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.model_selection import GroupKFold\n",
    "from sklearn.metrics import mean_squared_error, mean_absolute_error\n",
    "import statsmodels.formula.api as smf\n",
    "\n",
    "import statsmodels.api as sm\n",
    "import statsmodels.formula.api as smf"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d45b8698-2c81-4810-ad87-21cb57a1c38a",
   "metadata": {},
   "source": [
    "## Load the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "3fb3d72f-9500-4de2-9cf3-d9abc6a8330d",
   "metadata": {},
   "outputs": [],
   "source": [
    "td = pd.read_csv('T1_trainingData.csv')\n",
    "manID = pd.read_excel('manID.xlsx')\n",
    "womanID = pd.read_excel('womanID.xlsx')\n",
    "\n",
    "# Create a DataFrame for gender mapping\n",
    "manID['gender'] = 0\n",
    "womanID['gender'] = 1\n",
    "\n",
    "# Combine the manID and womanID into a single DataFrame\n",
    "gender_mapping = pd.concat([manID, womanID], ignore_index=True)\n",
    "\n",
    "# Ensure that the gender mapping DataFrame only contains ID and gender columns\n",
    "gender_mapping = gender_mapping[['ID', 'gender']]\n",
    "\n",
    "# Merge the gender mapping with td based on the ID\n",
    "gender_mapping['ID'] = gender_mapping['ID'].str.replace('_', '')\n",
    "gender_mapping['ID'] = pd.to_numeric(gender_mapping['ID'], errors='coerce')\n",
    "gender_mapping\n",
    "td = td.merge(gender_mapping, on='ID', how='left')\n",
    "\n",
    "# For any IDs that were not found in the gender mapping, fill with 'Unknown'\n",
    "td['gender'] = td['gender'].fillna('Unknown')\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "456e7076-27b5-419b-872a-429d8a4fe5dd",
   "metadata": {},
   "source": [
    "## Mixed learning Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b7675873-829a-4b76-b1cb-0f30fc719fa0",
   "metadata": {},
   "outputs": [],
   "source": [
    "#NORMALIZATION WITH MINMAX\n",
    "# Initialize the MinMaxScaler\n",
    "scaler_features = MinMaxScaler()\n",
    "scaler_target = MinMaxScaler()\n",
    "\n",
    "# Normalize the features 'F1' and 'F2'\n",
    "td[['F1', 'HNR' , 'Shimmer' , 'F2' , 'Jitter']] = scaler_feature.fit_transform(td[['F1', 'HNR' , 'Shimmer' , 'F2' , 'Jitter']])\n",
    "td[['UPDRS3_vf']] = scaler_target.fit_transform(td[['UPDRS3_vf']])\n",
    "\n",
    "\n",
    "# Fit the model  + Jitter + F2\n",
    "model = smf.mixedlm(\"UPDRS3_vf ~ Shimmer + Jitter + HNR + F1 + F2\", td, groups=td[\"ID\"])\n",
    "fitted_model = model.fit()\n",
    "print(fitted_model.summary())\n",
    "print(fitted_model.conf_int())\n",
    "\n",
    "# Predictions and actual values\n",
    "y_true = td['UPDRS3_vf']\n",
    "y_pred = fitted_model.fittedvalues\n",
    "\n",
    "mae = mean_absolute_error(y_true, y_pred)\n",
    "print(\"Mean Absolute Error (MAE):\", mae)\n",
    "\n",
    "epsilon = 1e-10\n",
    "# Relative Absolute Error Calculation\n",
    "relative_absolute_error = np.mean(np.abs(y_true - y_pred) / (np.abs(y_true) + epsilon))\n",
    "\n",
    "# Cross-Validation setup with GroupKFold (5 folds)\n",
    "group_kfold = GroupKFold(n_splits=5)\n",
    "mse_cv = []\n",
    "relative_absolute_error_cv = []\n",
    "maerror = []\n",
    "\n",
    "for train_index, test_index in group_kfold.split(td, groups=td['ID']):\n",
    "    train_data, test_data = td.iloc[train_index], td.iloc[test_index]\n",
    "    \n",
    "    model_cv = smf.mixedlm(\"UPDRS3_vf ~ Shimmer + Jitter + HNR + F1 + F2 \", train_data, groups=train_data[\"ID\"])\n",
    "    fitted_model_cv = model_cv.fit()\n",
    "    y_true_cv = test_data['UPDRS3_vf']\n",
    "    y_pred_cv = fitted_model_cv.predict(test_data)\n",
    "    \n",
    "    mse_cv.append(mean_squared_error(y_true_cv, y_pred_cv))\n",
    "    maerror.append(mean_absolute_error(y_true_cv, y_pred_cv))\n",
    "    relative_absolute_error_cv.append(np.mean(np.abs(y_true_cv - y_pred_cv) / (np.abs(y_true_cv) + epsilon)))\n",
    "\n",
    "# Mean of cross-validated metrics\n",
    "mse_cv_mean = np.mean(mse_cv)\n",
    "relative_absolute_error_cv_mean = np.mean(relative_absolute_error_cv)\n",
    "maerror_normalized  = np.mean(maerror)\n",
    "\n",
    "maerror_denormalized = maerror_normalized * (scaler_target.data_max_[0] - scaler_target.data_min_[0]) + scaler_target.data_min_[0]\n",
    "#maerror_denormalized = scaler_target.inverse_transform([[maerror_normalized]])[0, 0]\n",
    "# P-value analysis\n",
    "p_values = fitted_model.pvalues\n",
    "\n",
    "# Collecting results\n",
    "results = {\n",
    "    \"Relative Absolute Error\": relative_absolute_error,\n",
    "    \"Cross-validated MSE (Mean)\": mse_cv_mean,\n",
    "    \"Cross-validated Relative Absolute Error (Mean)\": relative_absolute_error_cv_mean,\n",
    "    \"P-values\": p_values,\n",
    "   \"Mean Absolute Error (MAE) (Normalized)\": maerror_normalized,\n",
    "    \"Mean Absolute Error (MAE) (Denormalized)\": maerror_denormalized\n",
    "}\n",
    "\n",
    "print(results)\n",
    "residuals = 0\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "residuals = fitted_model.resid\n",
    "plt.hist(residuals, bins=30)\n",
    "plt.title('Residuals Histogram')\n",
    "plt.xlabel('Residuals')\n",
    "plt.ylabel('Frequency')\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}