{ "cells": [ { "cell_type": "markdown", "id": "af3e8ee5", "metadata": {}, "source": [ "# Diabetes Classification + PCA" ] }, { "cell_type": "code", "execution_count": 1, "id": "e20e1efb", "metadata": {}, "outputs": [], "source": [ "\n", "import warnings\n", "warnings.filterwarnings(\"ignore\")\n", "\n", "import numpy as np\n", "import pandas as pd\n", "import matplotlib.pyplot as plt\n", "\n", "from sklearn.model_selection import train_test_split\n", "from sklearn.preprocessing import StandardScaler\n", "from sklearn.decomposition import PCA\n", "from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score, roc_auc_score\n", "\n", "import statsmodels.api as sm\n" ] }, { "cell_type": "code", "execution_count": 2, "id": "4e491137", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Размер: (768, 9)\n", "- Number of times pregnant\n", "- Plasma glucose concentration a 2 hours in an oral glucose tolerance test\n", "- Diastolic blood pressure (mm Hg)\n", "- Triceps skin fold thickness (mm)\n", "- 2-Hour serum insulin (mu U/ml)\n", "- Body mass index (weight in kg/(height in m)^2)\n", "- Diabetes pedigree function\n", "- Age (years)\n", "- Class variable (0 or 1)\n" ] }, { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
Number of times pregnantPlasma glucose concentration a 2 hours in an oral glucose tolerance testDiastolic blood pressure (mm Hg)Triceps skin fold thickness (mm)2-Hour serum insulin (mu U/ml)Body mass index (weight in kg/(height in m)^2)Diabetes pedigree functionAge (years)Class variable (0 or 1)
061487235033.60.627501
11856629026.60.351310
28183640023.30.672321
318966239428.10.167210
40137403516843.12.288331
\n", "
" ], "text/plain": [ " Number of times pregnant \\\n", "0 6 \n", "1 1 \n", "2 8 \n", "3 1 \n", "4 0 \n", "\n", " Plasma glucose concentration a 2 hours in an oral glucose tolerance test \\\n", "0 148 \n", "1 85 \n", "2 183 \n", "3 89 \n", "4 137 \n", "\n", " Diastolic blood pressure (mm Hg) Triceps skin fold thickness (mm) \\\n", "0 72 35 \n", "1 66 29 \n", "2 64 0 \n", "3 66 23 \n", "4 40 35 \n", "\n", " 2-Hour serum insulin (mu U/ml) \\\n", "0 0 \n", "1 0 \n", "2 0 \n", "3 94 \n", "4 168 \n", "\n", " Body mass index (weight in kg/(height in m)^2) Diabetes pedigree function \\\n", "0 33.6 0.627 \n", "1 26.6 0.351 \n", "2 23.3 0.672 \n", "3 28.1 0.167 \n", "4 43.1 2.288 \n", "\n", " Age (years) Class variable (0 or 1) \n", "0 50 1 \n", "1 31 0 \n", "2 32 1 \n", "3 21 0 \n", "4 33 1 " ] }, "execution_count": null, "metadata": {}, "output_type": "execute_result" } ], "source": [ "\n", "df = pd.read_csv(\"Diabetes Binary Classification.csv\")\n", "df.columns = df.columns.str.strip()\n", "\n", "print(\"Размер:\", df.shape)\n", "for c in df.columns:\n", " print(\"-\", c)\n", "df.head()\n" ] }, { "cell_type": "code", "execution_count": 3, "id": "eac0bf0b", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Train: (614, 8)\n", "Test: (154, 8)\n" ] } ], "source": [ "\n", "target = \"Class variable (0 or 1)\"\n", "\n", "numeric_cols = df.select_dtypes(include=[\"number\"]).columns.tolist()\n", "feature_cols = [c for c in numeric_cols if c != target]\n", "\n", "data = df[feature_cols + [target]].dropna().copy()\n", "\n", "X = data[feature_cols]\n", "y = data[target]\n", "\n", "X_train, X_test, y_train, y_test = train_test_split(\n", " X, y, test_size=0.2, random_state=42, stratify=y\n", ")\n", "\n", "print(\"Train:\", X_train.shape)\n", "print(\"Test:\", X_test.shape)\n" ] }, { "cell_type": "code", "execution_count": 4, "id": "74a40b66", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ " Logit Regression Results \n", "===================================================================================\n", "Dep. Variable: Class variable (0 or 1) No. Observations: 614\n", "Model: Logit Df Residuals: 605\n", "Method: MLE Df Model: 8\n", "Date: Sat, 14 Mar 2026 Pseudo R-squ.: 0.2789\n", "Time: 13:18:53 Log-Likelihood: -286.25\n", "converged: True LL-Null: -396.97\n", "Covariance Type: nonrobust LLR p-value: 1.909e-43\n", "============================================================================================================================================\n", " coef std err z P>|z| [0.025 0.975]\n", "--------------------------------------------------------------------------------------------------------------------------------------------\n", "const -0.8807 0.109 -8.064 0.000 -1.095 -0.667\n", "Number of times pregnant 0.3805 0.122 3.129 0.002 0.142 0.619\n", "Plasma glucose concentration a 2 hours in an oral glucose tolerance test 1.1661 0.135 8.609 0.000 0.901 1.432\n", "Diastolic blood pressure (mm Hg) -0.2048 0.110 -1.855 0.064 -0.421 0.012\n", "Triceps skin fold thickness (mm) 0.0695 0.126 0.551 0.581 -0.178 0.317\n", "2-Hour serum insulin (mu U/ml) -0.1371 0.124 -1.107 0.268 -0.380 0.106\n", "Body mass index (weight in kg/(height in m)^2) 0.7274 0.135 5.405 0.000 0.464 0.991\n", "Diabetes pedigree function 0.2593 0.111 2.335 0.020 0.042 0.477\n", "Age (years) 0.1819 0.121 1.499 0.134 -0.056 0.420\n", "============================================================================================================================================\n" ] } ], "source": [ "\n", "scaler = StandardScaler()\n", "X_train_scaled = scaler.fit_transform(X_train)\n", "X_test_scaled = scaler.transform(X_test)\n", "\n", "X_train_sm = sm.add_constant(\n", " pd.DataFrame(X_train_scaled, columns=feature_cols, index=y_train.index),\n", " has_constant=\"add\"\n", ")\n", "X_test_sm = sm.add_constant(\n", " pd.DataFrame(X_test_scaled, columns=feature_cols, index=y_test.index),\n", " has_constant=\"add\"\n", ")\n", "\n", "model = sm.Logit(y_train, X_train_sm).fit(disp=False)\n", "print(model.summary())\n" ] }, { "cell_type": "code", "execution_count": 5, "id": "a0143bf4", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Accuracy_test: 0.7142857142857143\n", "Precision_test: 0.6086956521739131\n", "Recall_test: 0.5185185185185185\n", "F1_test: 0.56\n", "ROC_AUC_test: 0.8240740740740741\n", "LLR_pvalue: 1.9087920276004937e-43\n", "Pseudo_R2: 0.2789142159398924\n" ] } ], "source": [ "\n", "train_proba = model.predict(X_train_sm)\n", "test_proba = model.predict(X_test_sm)\n", "\n", "train_pred = (train_proba >= 0.5).astype(int)\n", "test_pred = (test_proba >= 0.5).astype(int)\n", "\n", "print(\"Accuracy_test:\", accuracy_score(y_test, test_pred))\n", "print(\"Precision_test:\", precision_score(y_test, test_pred, zero_division=0))\n", "print(\"Recall_test:\", recall_score(y_test, test_pred, zero_division=0))\n", "print(\"F1_test:\", f1_score(y_test, test_pred, zero_division=0))\n", "print(\"ROC_AUC_test:\", roc_auc_score(y_test, test_proba))\n", "print(\"LLR_pvalue:\", model.llr_pvalue)\n", "print(\"Pseudo_R2:\", model.prsquared)\n" ] }, { "cell_type": "code", "execution_count": 6, "id": "3bed319c", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Explained variance 2PC: 0.47819876073856576\n" ] }, { "data": { "text/html": [ "
\n", "\n", "\n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", " \n", "
PC1PC2
Number of times pregnant0.1284320.593786
Plasma glucose concentration a 2 hours in an oral glucose tolerance test0.3930830.174029
Diastolic blood pressure (mm Hg)0.3600030.183892
Triceps skin fold thickness (mm)0.439824-0.331965
2-Hour serum insulin (mu U/ml)0.435026-0.250781
Body mass index (weight in kg/(height in m)^2)0.451941-0.100960
Diabetes pedigree function0.270611-0.122069
Age (years)0.1980270.620589
\n", "
" ], "text/plain": [ " PC1 PC2\n", "Number of times pregnant 0.128432 0.593786\n", "Plasma glucose concentration a 2 hours in an or... 0.393083 0.174029\n", "Diastolic blood pressure (mm Hg) 0.360003 0.183892\n", "Triceps skin fold thickness (mm) 0.439824 -0.331965\n", "2-Hour serum insulin (mu U/ml) 0.435026 -0.250781\n", "Body mass index (weight in kg/(height in m)^2) 0.451941 -0.100960\n", "Diabetes pedigree function 0.270611 -0.122069\n", "Age (years) 0.198027 0.620589" ] }, "metadata": {}, "output_type": "display_data" } ], "source": [ "\n", "X_scaled_all = StandardScaler().fit_transform(X)\n", "\n", "pca = PCA(n_components=2)\n", "X_pca = pca.fit_transform(X_scaled_all)\n", "\n", "print(\"Explained variance 2PC:\", pca.explained_variance_ratio_.sum())\n", "\n", "loadings = pd.DataFrame(\n", " pca.components_.T,\n", " index=feature_cols,\n", " columns=[\"PC1\", \"PC2\"]\n", ")\n", "display(loadings)\n" ] }, { "cell_type": "code", "execution_count": 7, "id": "594d79b5", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ " Logit Regression Results \n", "===================================================================================\n", "Dep. Variable: Class variable (0 or 1) No. Observations: 614\n", "Model: Logit Df Residuals: 611\n", "Method: MLE Df Model: 2\n", "Date: Sat, 14 Mar 2026 Pseudo R-squ.: 0.1749\n", "Time: 13:18:53 Log-Likelihood: -327.53\n", "converged: True LL-Null: -396.97\n", "Covariance Type: nonrobust LLR p-value: 6.954e-31\n", "==============================================================================\n", " coef std err z P>|z| [0.025 0.975]\n", "------------------------------------------------------------------------------\n", "const -0.7901 0.099 -7.979 0.000 -0.984 -0.596\n", "PC1 0.7629 0.081 9.418 0.000 0.604 0.922\n", "PC2 0.3654 0.071 5.155 0.000 0.226 0.504\n", "==============================================================================\n" ] } ], "source": [ "\n", "pca_df = pd.DataFrame(X_pca, columns=[\"PC1\", \"PC2\"], index=y.index)\n", "\n", "Xp_train, Xp_test, yp_train, yp_test = train_test_split(\n", " pca_df, y, test_size=0.2, random_state=42, stratify=y\n", ")\n", "\n", "Xp_train_sm = sm.add_constant(Xp_train, has_constant=\"add\")\n", "Xp_test_sm = sm.add_constant(Xp_test, has_constant=\"add\")\n", "\n", "model_pca = sm.Logit(yp_train, Xp_train_sm).fit(disp=False)\n", "print(model_pca.summary())\n" ] }, { "cell_type": "code", "execution_count": 8, "id": "b2b523ba", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Accuracy_test_pca: 0.7077922077922078\n", "Precision_test_pca: 0.6\n", "Recall_test_pca: 0.5\n", "F1_test_pca: 0.5454545454545454\n", "ROC_AUC_test_pca: 0.7088888888888889\n", "LLR_pvalue_pca: 6.95388111339125e-31\n", "Pseudo_R2_pca: 0.17492619879836246\n" ] } ], "source": [ "\n", "proba_pca = model_pca.predict(Xp_test_sm)\n", "pred_pca = (proba_pca >= 0.5).astype(int)\n", "\n", "print(\"Accuracy_test_pca:\", accuracy_score(yp_test, pred_pca))\n", "print(\"Precision_test_pca:\", precision_score(yp_test, pred_pca, zero_division=0))\n", "print(\"Recall_test_pca:\", recall_score(yp_test, pred_pca, zero_division=0))\n", "print(\"F1_test_pca:\", f1_score(yp_test, pred_pca, zero_division=0))\n", "print(\"ROC_AUC_test_pca:\", roc_auc_score(yp_test, proba_pca))\n", "print(\"LLR_pvalue_pca:\", model_pca.llr_pvalue)\n", "print(\"Pseudo_R2_pca:\", model_pca.prsquared)\n" ] }, { "cell_type": "markdown", "id": "a97c1b1e", "metadata": {}, "source": [ "## Итог\n", "\n", "\n", "Зависимая переменная — Class variable (0 or 1).\n", "\n", "Классификация по исходным признакам\n", "\n", "Качество модели:\n", "\n", "Accuracy = 0.714\n", "\n", "F1 = 0.56\n", "\n", "ROC-AUC = 0.824\n", "\n", "Pseudo R² = 0.279\n", "\n", "LLR p-value < 0.001\n", "\n", "Модель статистически значима. Наиболее значимые признаки:\n", "Plasma glucose, BMI, Number of pregnancies, Diabetes pedigree function.\n", "\n", "Снижение размерности (PCA)\n", "\n", "Две компоненты объясняют 47.8% дисперсии.\n", "\n", "Классификация по PC1 и PC2:\n", "\n", "Accuracy = 0.708\n", "\n", "F1 = 0.545\n", "\n", "ROC-AUC = 0.709\n", "\n", "Pseudo R² = 0.175\n", "\n", "Модель также значима, но качество ниже.\n", "\n", "Вывод\n", "\n", "Классификатор по исходным признакам лучше, так как имеет более высокий ROC-AUC и Pseudo R².\n", "Снижение размерности до двух компонент упрощает модель, но приводит к потере части информации и небольшому снижению точности. " ] } ], "metadata": { "language_info": { "name": "python" } }, "nbformat": 4, "nbformat_minor": 5 }