{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 作物病害分类模型训练示例\n",
    "\n",
    "本 Notebook 演示如何使用作物数据集训练机器学习模型，完成病害严重程度分类任务。\n",
    "\n",
    "**场景：** AI 模型训练（分类任务）\n",
    "\n",
    "**数据集：** CSV 作物病害标注表\n",
    "\n",
    "**模型：** 随机森林 / 决策树分类器"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1. 环境准备与数据加载"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.tree import DecisionTreeClassifier\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "from sklearn.metrics import (\n",
    "    classification_report, confusion_matrix, accuracy_score,\n",
    "    ConfusionMatrixDisplay\n",
    ")\n",
    "import os\n",
    "\n",
    "matplotlib.rcParams['font.sans-serif'] = ['SimHei', 'WenQuanYi Micro Hei']\n",
    "matplotlib.rcParams['axes.unicode_minus'] = False\n",
    "\n",
    "DATA_DIR = os.path.dirname(os.path.abspath(''))\n",
    "print('环境准备完成')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 加载数据\n",
    "df = pd.read_csv(f'{DATA_DIR}/csv/作物病害标注表.csv')\n",
    "print(f'数据集大小: {df.shape}')\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. 数据预处理"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 特征工程\n",
    "le_crop = LabelEncoder()\n",
    "le_disease = LabelEncoder()\n",
    "le_part = LabelEncoder()\n",
    "le_region = LabelEncoder()\n",
    "le_severity = LabelEncoder()\n",
    "\n",
    "df['作物编码'] = le_crop.fit_transform(df['作物'])\n",
    "df['病害编码'] = le_disease.fit_transform(df['病害名称'])\n",
    "df['部位编码'] = le_part.fit_transform(df['发病部位'])\n",
    "df['地区编码'] = le_region.fit_transform(df['地区'])\n",
    "df['严重程度编码'] = le_severity.fit_transform(df['严重程度'])\n",
    "\n",
    "# 选择特征\n",
    "feature_cols = ['作物编码', '病害编码', '部位编码', '地区编码', '温度_℃', '湿度_%', '经度', '纬度']\n",
    "X = df[feature_cols]\n",
    "y = df['严重程度编码']\n",
    "\n",
    "print(f'特征维度: {X.shape}')\n",
    "print(f'标签分布:\\n{pd.Series(y).value_counts().sort_index()}')\n",
    "print(f'\\n标签映射:')\n",
    "for i, cls in enumerate(le_severity.classes_):\n",
    "    print(f'  {i}: {cls}')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 3. 划分训练集和测试集"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "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",
    "print(f'训练集大小: {X_train.shape[0]}')\n",
    "print(f'测试集大小: {X_test.shape[0]}')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 4. 训练随机森林模型"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 训练随机森林\n",
    "rf_model = RandomForestClassifier(\n",
    "    n_estimators=100,\n",
    "    max_depth=10,\n",
    "    random_state=42,\n",
    "    n_jobs=-1\n",
    ")\n",
    "rf_model.fit(X_train, y_train)\n",
    "\n",
    "# 预测\n",
    "y_pred_rf = rf_model.predict(X_test)\n",
    "acc_rf = accuracy_score(y_test, y_pred_rf)\n",
    "print(f'随机森林准确率: {acc_rf:.4f}')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 5. 训练决策树模型（对比）"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 训练决策树\n",
    "dt_model = DecisionTreeClassifier(max_depth=8, random_state=42)\n",
    "dt_model.fit(X_train, y_train)\n",
    "\n",
    "y_pred_dt = dt_model.predict(X_test)\n",
    "acc_dt = accuracy_score(y_test, y_pred_dt)\n",
    "print(f'决策树准确率: {acc_dt:.4f}')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 6. 模型评估"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 分类报告\n",
    "print('=' * 50)\n",
    "print('随机森林 - 分类报告')\n",
    "print('=' * 50)\n",
    "print(classification_report(y_test, y_pred_rf, target_names=le_severity.classes_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 混淆矩阵可视化\n",
    "fig, axes = plt.subplots(1, 2, figsize=(14, 5))\n",
    "\n",
    "ConfusionMatrixDisplay.from_predictions(\n",
    "    y_test, y_pred_rf,\n",
    "    display_labels=le_severity.classes_,\n",
    "    cmap='Blues', ax=axes[0]\n",
    ")\n",
    "axes[0].set_title(f'随机森林混淆矩阵 (准确率={acc_rf:.4f})')\n",
    "\n",
    "ConfusionMatrixDisplay.from_predictions(\n",
    "    y_test, y_pred_dt,\n",
    "    display_labels=le_severity.classes_,\n",
    "    cmap='Oranges', ax=axes[1]\n",
    ")\n",
    "axes[1].set_title(f'决策树混淆矩阵 (准确率={acc_dt:.4f})')\n",
    "\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 7. 特征重要性分析"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 特征重要性\n",
    "importances = rf_model.feature_importances_\n",
    "indices = np.argsort(importances)[::-1]\n",
    "\n",
    "plt.figure(figsize=(10, 6))\n",
    "plt.bar(range(len(importances)), importances[indices], align='center', color='steelblue')\n",
    "plt.xticks(range(len(importances)), [feature_cols[i] for i in indices], rotation=45)\n",
    "plt.ylabel('重要性分数')\n",
    "plt.title('随机森林 - 特征重要性排序')\n",
    "plt.tight_layout()\n",
    "plt.show()\n",
    "\n",
    "print('特征重要性排名:')\n",
    "for rank, idx in enumerate(indices, 1):\n",
    "    print(f'  {rank}. {feature_cols[idx]}: {importances[idx]:.4f}')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 8. 模型预测示例"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 模拟新样本预测\n",
    "new_samples = pd.DataFrame({\n",
    "    '作物编码': le_crop.transform(['水稻', '小麦', '玉米']),\n",
    "    '病害编码': le_disease.transform(['稻瘟病', '锈病', '大斑病']),\n",
    "    '部位编码': le_part.transform(['叶片', '叶片', '穗部']),\n",
    "    '地区编码': le_region.transform(['长江中下游', '华北平原', '东北平原']),\n",
    "    '温度_℃': [28.5, 22.0, 25.3],\n",
    "    '湿度_%': [85.0, 65.0, 78.0],\n",
    "    '经度': [116.4, 114.5, 123.4],\n",
    "    '纬度': [30.5, 38.0, 41.8],\n",
    "})\n",
    "\n",
    "predictions = rf_model.predict(new_samples)\n",
    "pred_labels = le_severity.inverse_transform(predictions)\n",
    "\n",
    "print('预测结果:')\n",
    "print('-' * 50)\n",
    "for i, (_, row) in enumerate(new_samples.iterrows()):\n",
    "    crop_name = le_crop.inverse_transform([row['作物编码']])[0]\n",
    "    disease = le_disease.inverse_transform([row['病害编码']])[0]\n",
    "    print(f'样本{i+1}: {crop_name}/{disease} => 预测严重程度: {pred_labels[i]}')n",
    "\n",
    "print('\\n模型训练和预测流程演示完成！')"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "name": "python",
   "version": "3.10.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}