怎样建移动网站互联网大学生创新创业项目计划书

怎样建移动网站,互联网大学生创新创业项目计划书,二次开发怎么弄,网站 营销型基于python构建的一个完整的PMC#xff08;Policy Measurement and Comparison#xff09;政策文本量化评估系统#xff0c;使用Streamlit UI。一、系统架构概览1. 核心架构分层1. 前端交互层 (Streamlit UI)├── 多页面导航系统└── 交互式表单和可视化2. 业务逻辑层├…基于python构建的一个完整的PMCPolicy Measurement and Comparison政策文本量化评估系统使用Streamlit UI。一、系统架构概览1.核心架构分层1. 前端交互层 (Streamlit UI) ├── 多页面导航系统 └── 交互式表单和可视化 2. 业务逻辑层 ├── ConfigurationManager (配置管理) ├── 文本预处理模块 ├── 特征提取模块 └── PMC模型计算模块 3. 数据存储层 ├── Session State (内存存储) └── JSON文件 (持久化配置)2.PMC模型理论基础系统基于PMC政策量化评估模型包含9个一级变量政策性质、时效、激励、领域、工具、操作性、评估、受众、主题多个二级变量每个一级变量下细分的具体指标权重系统支持用户自定义各级权重二、核心模块详解1.配置管理系统 (ConfigurationManager)# 关键特性 - 完全可配置无预设词汇全部由用户定义 - 支持6种配置类型停用词、情感词、关键词库、权重、评估标准 - 导入/导出功能ZIP格式打包所有配置 - Session State管理实时保存用户配置2.文本处理流水线文本分析流程 1. 预处理 (FullyConfigurableTextPreprocessor) ├── 中文分词 (jieba) ├── 停用词过滤 ├── 情感分析基于用户词库 └── 可读性计算 2. 特征提取 (FullyConfigurablePolicyFeatureExtractor) ├── 关键词匹配 ├── 日期提取 ├── 情感倾向判断 └── 结构特征识别 3. PMC评分建议 └── 基于特征密度自动生成初步评分3.PMC计算模型评分公式 PMC指数 Σ(一级变量得分 × 权重) × 9 一级变量得分 Σ(二级变量得分 × 二级权重)三、页面功能解析1. 系统首页仪表盘展示关键指标配置状态检查快速开始指南2. 政策管理# 三种添加方式 1. 手动输入完整表单录入 2. 文件上传 - CSV格式结构化数据 - TXT格式批量处理 3. 文本粘贴支持JSON和纯文本3. 词汇配置核心特色停用词库过滤无关词汇情感词库正向/负向词汇定义特征关键词库JSON结构化的关键词分类配置持久化本地文件存储4. PMC评估# 交互式评分流程 1. 自动建议基于关键词匹配生成初步评分 2. 人工调整滑块手动调整0-1分值 3. 实时计算立即显示PMC指数和等级 4. 可视化雷达图展示各维度得分5. 可视化分析3D曲面图PMC得分的空间分布对比分析多政策雷达图对比热力图变量得分矩阵相关性分析变量间关联度6.⚙️ 高级配置权重调整一级变量权重配置总和为1评估标准自定义评分等级阈值系统管理会话保存/恢复四、技术亮点1.完全可配置的设计理念# 与传统系统的区别 传统系统预定义词典 → 用户适应系统 本系统用户自定义词典 → 系统适应业务2.智能特征映射系统自动将文本特征映射到PMC变量特征关键词库结构示例 { 预测性: [预测, 预计, 预期], 监管性: [监管, 管理, 监督], 人才引进: [人才, 引进, 培养] }3.多模态数据输入支持结构化CSV非结构化TXTJSON格式纯文本粘贴文件批量上传4.实时可视化Plotly图表交互式3D可视化Streamlit组件原生UI控件响应式设计自动适应布局五、应用场景1.政策研究机构批量评估政策文本质量多地区政策对比分析政策演化趋势研究2.政府部门政策草案自评估政策效果预评估跨部门政策协调性分析3.咨询公司为客户提供政策评估报告行业政策影响分析投资环境评估4.学术研究政策文本量化分析政策工具研究比较政策分析六、后期扩展1.可添加的功能1. 机器学习增强 - 自动关键词提取 - 主题模型分析 - 情感分析模型集成 2. 协作功能 - 多用户评估 - 专家评审系统 - 版本控制 3. 报告生成 - 自动生成评估报告 - PDF/Word导出 - 数据可视化报告2.性能优化异步处理大批量文本分布式计算支持缓存机制七、技术栈优势Streamlit快速原型开发交互性强Plotly丰富的可视化选项jieba中文分词准确率高JSON配置灵活可扩展模块化设计便于维护和扩展完整代码import streamlit as st import pandas as pd import numpy as np import plotly.graph_objects as go import plotly.express as px from datetime import datetime, timedelta import json import re import os import sys import warnings from collections import defaultdict, Counter from typing import Dict, List, Tuple, Any, Optional, Set import jieba import jieba.analyse import tempfile import hashlib import zipfile warnings.filterwarnings(ignore) # 设置页面配置 st.set_page_config( page_titlePMC政策文本量化评估系统, page_icon⚙️, layoutwide, initial_sidebar_stateexpanded ) # 应用标题 st.title(⚙️ PMC政策文本量化评估系统) st.markdown( **完全可配置的政策文本量化评估工具**所有词汇表均由用户自行配置无预设默认词汇。 支持 - 用户自定义所有词汇表停用词、情感词、关键词库等 - 支持TXT、CSV、JSON格式上传 - 政策文本自动预处理与特征提取 - PMC指数模型计算与可视化分析 - 批量政策对比评估 ) # 配置管理器 class ConfigurationManager: 配置管理器处理用户自定义词汇表 def __init__(self): self.config_dir ./configs os.makedirs(self.config_dir, exist_okTrue) def save_user_config(self, config_type: str, data: Any, config_name: str user_config): 保存用户配置 config_path os.path.join(self.config_dir, f{config_name}_{config_type}.json) if config_type stopwords: data list(data) if isinstance(data, set) else data elif config_type feature_keywords: if not isinstance(data, dict): raise ValueError(特征关键词库必须是字典格式) with open(config_path, w, encodingutf-8) as f: json.dump(data, f, ensure_asciiFalse, indent2) return config_path def load_user_config(self, config_type: str, config_name: str user_config): 加载用户配置 config_path os.path.join(self.config_dir, f{config_name}_{config_type}.json) if os.path.exists(config_path): with open(config_path, r, encodingutf-8) as f: data json.load(f) if config_type stopwords: data set(data) return data else: return None def export_all_configs(self, config_name: str pmc_config): 导出所有配置为ZIP文件 import zipfile temp_dir tempfile.mkdtemp() zip_path os.path.join(temp_dir, f{config_name}.zip) with zipfile.ZipFile(zip_path, w) as zipf: config_types [stopwords, positive_words, negative_words, feature_keywords, pmc_weights, evaluation_criteria] for config_type in config_types: config_data self.load_user_config(config_type, config_name) if config_data is None: # 没有用户配置创建空的配置文件 if config_type stopwords: config_data [] elif config_type in [positive_words, negative_words]: config_data [] elif config_type feature_keywords: config_data {} elif config_type pmc_weights: config_data {primary_weights: {}, secondary_weights: {}} elif config_type evaluation_criteria: config_data {levels: {}} temp_file os.path.join(temp_dir, f{config_type}.json) with open(temp_file, w, encodingutf-8) as f: json.dump(config_data, f, ensure_asciiFalse, indent2) zipf.write(temp_file, f{config_type}.json) return zip_path def import_configs_from_zip(self, zip_file): 从ZIP文件导入配置 import zipfile temp_dir tempfile.mkdtemp() with zipfile.ZipFile(zip_file, r) as zipf: zipf.extractall(temp_dir) configs {} config_types [stopwords, positive_words, negative_words, feature_keywords, pmc_weights, evaluation_criteria] for config_type in config_types: config_file os.path.join(temp_dir, f{config_type}.json) if os.path.exists(config_file): with open(config_file, r, encodingutf-8) as f: configs[config_type] json.load(f) return configs # 核心模型类 class FullyConfigurableTextPreprocessor: 完全可配置的文本预处理器 def __init__(self, config_managerNone): self.config_manager config_manager or ConfigurationManager() # 加载配置 self.stopwords self._load_stopwords() self.positive_words self._load_positive_words() self.negative_words self._load_negative_words() # 初始化jieba try: jieba.initialize() except: pass def _load_stopwords(self) - Set[str]: 加载停用词表完全由用户配置 # 先尝试加载用户配置 user_stopwords self.config_manager.load_user_config(stopwords) if user_stopwords is not None: return set(user_stopwords) # 检查session state中是否有上传的停用词 if user_stopwords in st.session_state and st.session_state.user_stopwords: user_words set(st.session_state.user_stopwords.split(\n)) return user_words # 无默认停用词返回空集合 return set() def _load_positive_words(self) - List[str]: 加载正面词汇完全由用户配置 # 先尝试加载用户配置 user_positive self.config_manager.load_user_config(positive_words) if user_positive is not None: return user_positive # 检查session state中是否有上传的正面词汇 if user_positive_words in st.session_state and st.session_state.user_positive_words: return [word.strip() for word in st.session_state.user_positive_words.split(\n) if word.strip()] # 无默认正面词汇返回空列表 return [] def _load_negative_words(self) - List[str]: 加载负面词汇完全由用户配置 # 先尝试加载用户配置 user_negative self.config_manager.load_user_config(negative_words) if user_negative is not None: return user_negative # 检查session state中是否有上传的负面词汇 if user_negative_words in st.session_state and st.session_state.user_negative_words: return [word.strip() for word in st.session_state.user_negative_words.split(\n) if word.strip()] # 无默认负面词汇返回空列表 return [] def update_stopwords(self, stopwords_text: str): 更新停用词表 stopwords_list [word.strip() for word in stopwords_text.split(\n) if word.strip()] st.session_state.user_stopwords stopwords_text self.stopwords set(stopwords_list) def update_sentiment_words(self, positive_text: str, negative_text: str): 更新情感词汇 positive_list [word.strip() for word in positive_text.split(\n) if word.strip()] negative_list [word.strip() for word in negative_text.split(\n) if word.strip()] st.session_state.user_positive_words positive_text st.session_state.user_negative_words negative_text self.positive_words positive_list self.negative_words negative_list def preprocess_text(self, text: str) - List[str]: 预处理政策文本 # 去除特殊字符和数字 text re.sub(r[^\u4e00-\u9fa5a-zA-Z], , text) # 分词 words jieba.lcut(text) # 去除停用词 words [word for word in words if word not in self.stopwords and len(word) 1] return words def extract_keywords(self, text: str, top_k: int 20) - List[Tuple[str, float]]: 提取关键词 keywords jieba.analyse.extract_tags(text, topKtop_k, withWeightTrue) return keywords def analyze_sentiment(self, text: str) - Dict[str, Any]: 分析文本情感 words self.preprocess_text(text) # 如果用户没有配置情感词返回中性结果 if not self.positive_words and not self.negative_words: return { sentiment_score: 0.5, positive_count: 0, negative_count: 0, sentiment: 中性 (未配置情感词) } pos_count sum(1 for word in words if word in self.positive_words) neg_count sum(1 for word in words if word in self.negative_words) if pos_count neg_count 0: sentiment_score 0.5 else: sentiment_score pos_count / (pos_count neg_count) sentiment 中性 if sentiment_score 0.6: sentiment 积极 elif sentiment_score 0.4: sentiment 消极 return { sentiment_score: sentiment_score, positive_count: pos_count, negative_count: neg_count, sentiment: sentiment } def extract_dates(self, text: str) - List[str]: 提取日期信息 date_patterns [ r\d{4}年\d{1,2}月\d{1,2}日, r\d{4}年\d{1,2}月, r\d{4}年, r\d{4}-\d{1,2}-\d{1,2}, r\d{4}至\d{4}年, r\d{4}—\d{4}年 ] dates [] for pattern in date_patterns: matches re.findall(pattern, text) dates.extend(matches) return dates def calculate_readability(self, text: str) - Dict[str, float]: 计算文本可读性指标 sentences re.split(r[。;.!?], text) sentences [s.strip() for s in sentences if s.strip()] words self.preprocess_text(text) if not sentences or not words: return {avg_sentence_length: 0, avg_word_length: 0, readability_score: 0} avg_sentence_length len(words) / len(sentences) avg_word_length sum(len(word) for word in words) / len(words) if words else 0 readability_score 100 - (avg_sentence_length * 0.5 avg_word_length * 0.3) return { avg_sentence_length: avg_sentence_length, avg_word_length: avg_word_length, readability_score: max(0, min(100, readability_score)) } class FullyConfigurablePolicyFeatureExtractor: 完全可配置的政策特征提取器 def __init__(self, config_managerNone): self.config_manager config_manager or ConfigurationManager() self.preprocessor FullyConfigurableTextPreprocessor(config_manager) # 加载特征关键词库 self.feature_keywords self._load_feature_keywords() def _load_feature_keywords(self) - Dict[str, List[str]]: 加载特征关键词库完全由用户配置 # 先尝试加载用户配置 user_keywords self.config_manager.load_user_config(feature_keywords) if user_keywords is not None: return user_keywords # 检查session state中是否有上传的关键词库 if user_feature_keywords in st.session_state and st.session_state.user_feature_keywords: try: keywords_dict json.loads(st.session_state.user_feature_keywords) if isinstance(keywords_dict, dict): return keywords_dict except: pass # 无默认关键词库返回空字典 return {} def update_feature_keywords(self, keywords_text: str): 更新特征关键词库 try: keywords_dict json.loads(keywords_text) if isinstance(keywords_dict, dict): st.session_state.user_feature_keywords keywords_text self.feature_keywords keywords_dict return True else: return False except: return False def extract_features(self, text: str) - Dict[str, Any]: 从政策文本中提取特征 words self.preprocess_text(text) word_count len(words) # 关键词匹配 keyword_features {} for category, keywords in self.feature_keywords.items(): matches [] for keyword in keywords: if keyword in text: matches.append(keyword) keyword_features[category] { count: len(matches), keywords: matches, density: len(matches) / word_count if word_count 0 else 0 } # 情感分析 sentiment self.preprocessor.analyze_sentiment(text) # 日期提取 dates self.preprocessor.extract_dates(text) # 可读性分析 readability self.preprocessor.calculate_readability(text) # 提取关键词 keywords self.preprocessor.extract_keywords(text, top_k15) return { basic_features: { word_count: word_count, unique_words: len(set(words)), dates_found: len(dates), dates: dates }, keyword_features: keyword_features, sentiment_analysis: sentiment, readability_analysis: readability, top_keywords: keywords } def suggest_pmc_scores(self, features: Dict[str, Any]) - Dict[str, Dict[str, float]]: 基于特征提取建议PMC评分 pmc_suggestions {} # 映射关系特征类别 - PMC变量 feature_to_pmc { 预测性: (X1, X1a_预测性), 监管性: (X1, X1b_监管性), 建议性: (X1, X1c_建议性), 人才引进: (X3, X3a_人才引进), 资金支持: (X3, X3b_资金支持), 税收优惠: (X3, X3c_税收优惠), 法律保障: (X3, X3d_法律保障), 基础设施建设: (X4, X4a_基础设施建设), 技术研发: (X4, X4b_技术研发), 产业培育: (X4, X4c_产业培育), 市场应用: (X4, X4d_市场应用), 供给型: (X5, X5a_供给型), 需求型: (X5, X5b_需求型), 环境型: (X5, X5c_环境型), 政府部门: (X8, X8a_政府部门), 企业: (X8, X8b_企业), 科研机构: (X8, X8c_科研机构), 个人: (X8, X8d_个人) } # 根据特征密度生成建议分数 for feature_category, pmc_info in feature_to_pmc.items(): primary_var, secondary_var pmc_info if primary_var not in pmc_suggestions: pmc_suggestions[primary_var] {} feature_data features[keyword_features].get(feature_category, {}) density feature_data.get(density, 0) # 基于密度计算分数0-1之间 score min(1.0, density * 50) pmc_suggestions[primary_var][secondary_var] round(score, 2) # 补充其他变量的默认值 default_scores { X2: {X2a_长期(5年): 0.5, X2b_中期(3-5年): 0.5, X2c_短期(3年): 0.5}, X6: {X6a_目标明确性: 0.5, X6b_任务具体性: 0.5, X6c_责任主体清晰性: 0.5}, X7: {X7a_评估指标: 0.5, X7b_评估方法: 0.5, X7c_评估周期: 0.5}, X9: {X9a_产业激励: 0.5, X9b_基建驱动: 0.5, X9c_场景应用: 0.5} } for var, scores in default_scores.items(): if var not in pmc_suggestions: pmc_suggestions[var] {} for sec_var, score in scores.items(): if sec_var not in pmc_suggestions[var]: pmc_suggestions[var][sec_var] score return pmc_suggestions class FullyConfigurablePMCModel: 完全可配置的PMC指数模型 def __init__(self, config_managerNone): self.config_manager config_manager or ConfigurationManager() # 定义PMC变量结构 self.primary_variables { X1: 政策性质, X2: 政策时效, X3: 政策激励, X4: 政策领域, X5: 政策工具, X6: 政策操作性, X7: 政策评估, X8: 政策受众, X9: 政策主题 } self.secondary_variables { X1: [X1a_预测性, X1b_监管性, X1c_建议性], X2: [X2a_长期(5年), X2b_中期(3-5年), X2c_短期(3年)], X3: [X3a_人才引进, X3b_资金支持, X3c_税收优惠, X3d_法律保障], X4: [X4a_基础设施建设, X4b_技术研发, X4c_产业培育, X4d_市场应用], X5: [X5a_供给型, X5b_需求型, X5c_环境型], X6: [X6a_目标明确性, X6b_任务具体性, X6c_责任主体清晰性], X7: [X7a_评估指标, X7b_评估方法, X7c_评估周期], X8: [X8a_政府部门, X8b_企业, X8c_科研机构, X8d_个人], X9: [X9a_产业激励, X9b_基建驱动, X9c_场景应用] } # 加载权重 self.weights self._load_weights() self.secondary_weights self._load_secondary_weights() # 加载评估标准 self.evaluation_criteria self._load_evaluation_criteria() # 初始化其他组件 self.feature_extractor FullyConfigurablePolicyFeatureExtractor(config_manager) self.preprocessor FullyConfigurableTextPreprocessor(config_manager) def _load_weights(self) - Dict[str, float]: 加载权重配置 user_weights self.config_manager.load_user_config(pmc_weights) if user_weights and primary_weights in user_weights: return user_weights[primary_weights] # 检查session state if user_pmc_weights in st.session_state and st.session_state.user_pmc_weights: try: weights_dict json.loads(st.session_state.user_pmc_weights) if isinstance(weights_dict, dict): # 验证权重总和为1 total sum(weights_dict.values()) if abs(total - 1.0) 0.01: # 允许小误差 return weights_dict except: pass # 无用户配置使用等权重 num_vars len(self.primary_variables) equal_weight 1.0 / num_vars return {var: equal_weight for var in self.primary_variables.keys()} def _load_secondary_weights(self) - Dict[str, List[float]]: 加载二级权重配置 user_weights self.config_manager.load_user_config(pmc_weights) if user_weights and secondary_weights in user_weights: return user_weights[secondary_weights] # 无用户配置使用等权重 secondary_weights {} for var, sec_vars in self.secondary_variables.items(): num_sec_vars len(sec_vars) secondary_weights[var] [1.0 / num_sec_vars] * num_sec_vars return secondary_weights def _load_evaluation_criteria(self) - Dict[str, Any]: 加载评估标准 user_criteria self.config_manager.load_user_config(evaluation_criteria) if user_criteria is not None: return user_criteria # 检查session state if user_evaluation_criteria in st.session_state and st.session_state.user_evaluation_criteria: try: criteria_dict json.loads(st.session_state.user_evaluation_criteria) if isinstance(criteria_dict, dict): return criteria_dict except: pass # 无用户配置使用默认评估标准 return { levels: { 优秀: {min: 8.0, max: 9.0, color: #4CAF50}, 良好: {min: 7.0, max: 8.0, color: #8BC34A}, 及格: {min: 6.0, max: 7.0, color: #FFC107}, 较差: {min: 0.0, max: 6.0, color: #F44336} } } def update_weights(self, weights_text: str) - bool: 更新权重配置 try: weights_dict json.loads(weights_text) if isinstance(weights_dict, dict): st.session_state.user_pmc_weights weights_text self.weights weights_dict return True else: return False except: return False def update_evaluation_criteria(self, criteria_text: str) - bool: 更新评估标准 try: criteria_dict json.loads(criteria_text) if isinstance(criteria_dict, dict): st.session_state.user_evaluation_criteria criteria_text self.evaluation_criteria criteria_dict return True else: return False except: return False def calculate_pmc_index(self, policy_data: Dict) - float: 计算PMC指数 total_score 0 for primary_var in self.primary_variables.keys(): if primary_var in policy_data: secondary_scores [] secondary_vars self.secondary_variables[primary_var] for i, sec_var in enumerate(secondary_vars): if sec_var in policy_data[primary_var]: weighted_score policy_data[primary_var][sec_var] * self.secondary_weights[primary_var][i] secondary_scores.append(weighted_score) if secondary_scores: primary_score sum(secondary_scores) weighted_primary primary_score * self.weights[primary_var] * 9 total_score weighted_primary return round(total_score, 2) def evaluate_policy_level(self, pmc_score: float) - Tuple[str, str]: 评估政策等级 levels self.evaluation_criteria.get(levels, { 优秀: {min: 8.0, max: 9.0, color: #4CAF50}, 良好: {min: 7.0, max: 8.0, color: #8BC34A}, 及格: {min: 6.0, max: 7.0, color: #FFC107}, 较差: {min: 0.0, max: 6.0, color: #F44336} }) for level_name, level_info in levels.items(): if level_info[min] pmc_score level_info[max]: return level_name, level_info[color] return 未知, #9E9E9E def calculate_variable_scores(self, policy_data: Dict) - Dict[str, float]: 计算每个一级变量的得分 variable_scores {} for primary_var in self.primary_variables.keys(): if primary_var in policy_data: secondary_scores [] secondary_vars self.secondary_variables[primary_var] for i, sec_var in enumerate(secondary_vars): if sec_var in policy_data[primary_var]: weighted_score policy_data[primary_var][sec_var] * self.secondary_weights[primary_var][i] secondary_scores.append(weighted_score) if secondary_scores: variable_scores[primary_var] sum(secondary_scores) return variable_scores # Streamlit应用 # 初始化session state if policies not in st.session_state: st.session_state.policies [] if pmc_results not in st.session_state: st.session_state.pmc_results {} if analysis_history not in st.session_state: st.session_state.analysis_history [] if config_manager not in st.session_state: st.session_state.config_manager ConfigurationManager() if pmc_model not in st.session_state: st.session_state.pmc_model FullyConfigurablePMCModel(st.session_state.config_manager) # 侧边栏 st.sidebar.title(⚙️ 系统配置) # 功能选择 page st.sidebar.radio( 导航菜单, [ 系统首页, 政策管理, 词汇配置, PMC评估, 可视化分析, ⚙️ 高级配置] ) # 首页 if page 系统首页: st.header(PMC政策文本量化评估系统) col1, col2, col3, col4 st.columns(4) with col1: policy_count len(st.session_state.policies) st.metric(政策总数, policy_count) with col2: evaluated_count len(st.session_state.pmc_results) st.metric(已评估政策, evaluated_count) with col3: if st.session_state.pmc_results: avg_score np.mean([v[pmc_score] for v in st.session_state.pmc_results.values()]) else: avg_score 0 st.metric(平均PMC指数, f{avg_score:.2f}) with col4: if st.session_state.pmc_results: levels [v.get(level, 未知) for v in st.session_state.pmc_results.values()] excellent_count levels.count(优秀) else: excellent_count 0 st.metric(优秀政策, excellent_count) st.markdown(---) # 系统介绍 st.subheader( 系统功能概述) features [ {title: 完全可配置词汇库, desc: 所有词汇表完全由用户配置无预设词汇, icon: }, {title: 灵活上传方式, desc: 支持文本输入、文件上传、批量处理, icon: }, {title: 智能文本分析, desc: 自动提取政策特征识别关键变量, icon: }, {title: 多投入产出表, desc: 构建标准化数据分析框架, icon: }, {title: PMC指数计算, desc: 科学量化政策质量, icon: }, {title: 多维可视化, desc: 雷达图、曲面图、热力图等多种展示, icon: } ] cols st.columns(3) for i, feature in enumerate(features): with cols[i % 3]: st.markdown(f div stylepadding: 20px; border-radius: 10px; background-color: #f8f9fa; margin-bottom: 10px; h4{feature[icon]} {feature[title]}/h4 p stylecolor: #666; font-size: 14px;{feature[desc]}/p /div , unsafe_allow_htmlTrue) # 快速开始指南 with st.expander( 快速开始指南, expandedTrue): st.markdown( 1. **配置词汇库**在词汇配置页面设置停用词、情感词、关键词库必须配置 2. **添加政策**在政策管理页面导入或添加政策文本 3. **文本分析**系统自动分析政策文本特征 4. **PMC评估**在PMC评估页面进行量化评估 5. **可视化**在可视化分析页面查看多维图表 6. **高级配置**在高级配置页面调整权重和评估标准 ) # 配置状态检查 st.subheader( 配置状态检查) config_status { 停用词配置: user_stopwords in st.session_state and st.session_state.user_stopwords, 正面词汇配置: user_positive_words in st.session_state and st.session_state.user_positive_words, 负面词汇配置: user_negative_words in st.session_state and st.session_state.user_negative_words, 特征关键词库配置: user_feature_keywords in st.session_state and st.session_state.user_feature_keywords, PMC权重配置: user_pmc_weights in st.session_state and st.session_state.user_pmc_weights, 评估标准配置: user_evaluation_criteria in st.session_state and st.session_state.user_evaluation_criteria } cols st.columns(3) for i, (config_name, configured) in enumerate(config_status.items()): with cols[i % 3]: if configured: st.success(f✅ {config_name}) else: st.warning(f⚠️ {config_name} (未配置)) if not all(config_status.values()): st.info( 请先完成所有词汇库的配置以获得最佳分析效果。) # 政策管理页面 elif page 政策管理: st.header(政策管理) tab1, tab2, tab3 st.tabs([ 添加政策, 政策列表, 批量导入]) with tab1: st.subheader(添加新政策) with st.form(add_policy_form): col1, col2 st.columns(2) with col1: policy_name st.text_input(政策名称*, placeholder例如低空经济发展指导意见) policy_level st.selectbox(政策层级*, [国家, 省级, 地市级, 区县级, 其他]) industry st.selectbox(所属行业*, [低空经济, 数字经济, 新能源, 人工智能, 生物医药, 高端制造, 其他]) with col2: policy_region st.text_input(适用地区*, placeholder例如全国、四川省、苏州市) policy_date st.date_input(发布日期, datetime.now()) policy_source st.text_input(政策来源, placeholder例如国务院、发改委) policy_description st.text_area(政策描述, placeholder简要描述政策的主要内容和目标, height100) policy_text st.text_area(政策全文*, placeholder请输入政策全文内容..., height300) submitted st.form_submit_button( 添加政策) if submitted: if not policy_name or not policy_text: st.error(请填写政策名称和全文) else: # 生成政策ID new_id max([p[id] for p in st.session_state.policies]) 1 if st.session_state.policies else 1 # 创建政策对象 new_policy { id: new_id, name: policy_name, level: policy_level, region: policy_region, industry: industry, date: policy_date.strftime(%Y-%m-%d), source: policy_source, description: policy_description, text: policy_text, created_at: datetime.now().isoformat(), hash: hashlib.md5(policy_text.encode()).hexdigest()[:16] } st.session_state.policies.append(new_policy) # 记录分析历史 st.session_state.analysis_history.append({ timestamp: datetime.now().isoformat(), action: 添加政策, policy_name: policy_name, policy_id: new_id }) st.success(f✅ 政策 {policy_name} 添加成功ID: {new_id}) # 自动分析文本 with st.spinner(正在自动分析政策文本...): try: feature_extractor FullyConfigurablePolicyFeatureExtractor(st.session_state.config_manager) features feature_extractor.extract_features(policy_text) suggestions feature_extractor.suggest_pmc_scores(features) pmc_score st.session_state.pmc_model.calculate_pmc_index(suggestions) level, color st.session_state.pmc_model.evaluate_policy_level(pmc_score) variable_scores st.session_state.pmc_model.calculate_variable_scores(suggestions) # 保存分析结果 st.session_state.pmc_results[new_id] { pmc_score: pmc_score, level: level, color: color, variable_scores: variable_scores, text_analysis: features, pmc_suggestions: suggestions } st.info(f 自动分析完成PMC指数: {pmc_score:.2f} ({level})) except Exception as e: st.warning(f文本分析过程中出现错误: {str(e)}) with tab2: st.subheader(政策列表) if st.session_state.policies: # 创建数据框 policy_data [] for policy in st.session_state.policies: pmc_info st.session_state.pmc_results.get(policy[id], {}) policy_data.append({ ID: policy[id], 政策名称: policy[name], 行业: policy.get(industry, ), 层级: policy[level], 地区: policy[region], 发布日期: policy[date], 字数: len(policy[text]), PMC指数: pmc_info.get(pmc_score, 未评估), 评估等级: pmc_info.get(level, 未评估) }) df pd.DataFrame(policy_data) # 搜索和筛选 col1, col2, col3 st.columns(3) with col1: search_name st.text_input(搜索政策名称) with col2: filter_industry st.selectbox(筛选行业, [全部] list(df[行业].unique())) with col3: filter_level st.selectbox(筛选层级, [全部] list(df[层级].unique())) # 应用筛选 filtered_df df if search_name: filtered_df filtered_df[filtered_df[政策名称].str.contains(search_name, naFalse)] if filter_industry ! 全部: filtered_df filtered_df[filtered_df[行业] filter_industry] if filter_level ! 全部: filtered_df filtered_df[filtered_df[层级] filter_level] # 显示表格 st.dataframe( filtered_df, use_container_widthTrue, column_config{ PMC指数: st.column_config.NumberColumn(format%.2f), 字数: st.column_config.NumberColumn(format%d) }, hide_indexTrue ) # 导出数据 if st.button( 导出政策数据): csv df.to_csv(indexFalse).encode(utf-8-sig) st.download_button( label下载CSV, datacsv, file_namepolicies_data.csv, mimetext/csv ) else: st.info(暂无政策数据请先添加政策。) with tab3: st.subheader(批量导入政策) upload_option st.radio(导入方式, [上传CSV文件, 上传TXT文件, 粘贴文本]) if upload_option 上传CSV文件: uploaded_file st.file_uploader(选择CSV文件, type[csv], helpCSV文件应包含以下列name, level, region, industry, date, source, description, text) if uploaded_file is not None: try: df pd.read_csv(uploaded_file) # 检查必要列 required_columns [name, text] if all(col in df.columns for col in required_columns): progress_bar st.progress(0) imported_count 0 for idx, row in df.iterrows(): new_id max( [p[id] for p in st.session_state.policies]) 1 if st.session_state.policies else 1 new_policy { id: new_id, name: row.get(name, f政策_{new_id}), level: row.get(level, 其他), region: row.get(region, 未知), industry: row.get(industry, 其他), date: row.get(date, datetime.now().strftime(%Y-%m-%d)), source: row.get(source, uploaded_file.name), description: row.get(description, ), text: str(row.get(text, ))[:5000], created_at: datetime.now().isoformat(), hash: hashlib.md5(str(row.get(text, )).encode()).hexdigest()[:16] } st.session_state.policies.append(new_policy) imported_count 1 progress_bar.progress((idx 1) / len(df)) st.success(f✅ 成功导入 {imported_count} 条政策) else: st.error(fCSV文件必须包含以下列{, .join(required_columns)}) except Exception as e: st.error(f文件解析失败: {str(e)}) elif upload_option 上传TXT文件: uploaded_files st.file_uploader(选择TXT文件, type[txt], accept_multiple_filesTrue) if uploaded_files: progress_bar st.progress(0) imported_count 0 for i, uploaded_file in enumerate(uploaded_files): try: content uploaded_file.getvalue().decode(utf-8) # 简单处理每段作为一个政策 paragraphs [p.strip() for p in content.split(\n\n) if p.strip()] for para in paragraphs: if len(para) 100: # 只处理较长的段落 new_id max([p[id] for p in st.session_state.policies]) 1 if st.session_state.policies else 1 new_policy { id: new_id, name: f{uploaded_file.name}_{new_id}, level: 其他, region: 未知, industry: 其他, date: datetime.now().strftime(%Y-%m-%d), source: uploaded_file.name, description: , text: para[:5000], created_at: datetime.now().isoformat(), hash: hashlib.md5(para.encode()).hexdigest()[:16] } st.session_state.policies.append(new_policy) imported_count 1 progress_bar.progress((i 1) / len(uploaded_files)) except Exception as e: st.warning(f文件 {uploaded_file.name} 处理失败: {str(e)}) if imported_count 0: st.success(f✅ 成功导入 {imported_count} 条政策) elif upload_option 粘贴文本: text_input st.text_area(粘贴政策文本每行一个政策或JSON格式, height200) col1, col2 st.columns(2) with col1: format_type st.selectbox(文本格式, [每行一个政策, JSON格式]) with col2: if st.button(解析文本): if text_input: if format_type 每行一个政策: policies [p.strip() for p in text_input.split(\n) if p.strip()] else: # JSON格式 try: data json.loads(text_input) if isinstance(data, list): policies data else: policies [data] except: st.error(JSON格式不正确) policies [] imported_count 0 for i, policy_text in enumerate(policies): if isinstance(policy_text, dict): # 如果是字典提取文本字段 text_content policy_text.get(text, ) if not text_content: continue else: text_content str(policy_text) if len(text_content) 50: new_id max([p[id] for p in st.session_state.policies]) 1 if st.session_state.policies else 1 new_policy { id: new_id, name: f粘贴政策_{new_id}, level: 其他, region: 未知, industry: 其他, date: datetime.now().strftime(%Y-%m-%d), source: 文本粘贴, description: , text: text_content[:5000], created_at: datetime.now().isoformat(), hash: hashlib.md5(text_content.encode()).hexdigest()[:16] } st.session_state.policies.append(new_policy) imported_count 1 st.success(f✅ 成功添加 {imported_count} 条政策) # 词汇配置页面 elif page 词汇配置: st.header(词汇库配置) st.markdown(**重要提示**本系统完全由用户配置无预设词汇。请先配置词汇库否则分析功能可能受限。) tab1, tab2, tab3, tab4 st.tabs([ 停用词配置, 情感词配置, 关键词库配置, 导入/导出配置]) with tab1: st.subheader(停用词配置) st.markdown(停用词在文本分析时会被过滤掉每行一个词) st.warning(⚠️ 未配置停用词可能导致文本分析不准确) # 获取当前停用词 current_stopwords if user_stopwords in st.session_state: current_stopwords st.session_state.user_stopwords col1, col2 st.columns([2, 1]) with col1: stopwords_input st.text_area( 停用词列表每行一个词, valuecurrent_stopwords, height300, help输入需要过滤的停用词每行一个。例如的、了、在、是、我、有、和、就、不、人 ) with col2: st.markdown(### 操作) if st.button( 保存停用词, typeprimary): if stopwords_input: # 保存到session state st.session_state.user_stopwords stopwords_input # 保存到文件 stopwords_list [word.strip() for word in stopwords_input.split(\n) if word.strip()] st.session_state.config_manager.save_user_config(stopwords, stopwords_list) st.success(✅ 停用词已保存) else: st.warning(请输入停用词) st.markdown(---) # 文件上传 uploaded_file st.file_uploader(上传停用词文件, type[txt]) if uploaded_file is not None: content uploaded_file.getvalue().decode(utf-8) st.text_area(文件内容预览, content[:1000], height150) if st.button( 使用上传文件): st.session_state.user_stopwords content stopwords_list [word.strip() for word in content.split(\n) if word.strip()] st.session_state.config_manager.save_user_config(stopwords, stopwords_list) st.success(✅ 已使用上传的停用词文件) st.rerun() st.markdown(---) if st.button(️ 清空停用词): st.session_state.user_stopwords st.success(✅ 已清空停用词) st.rerun() # 显示统计信息 if stopwords_input: stopwords_count len([word for word in stopwords_input.split(\n) if word.strip()]) st.info(f当前停用词数量: {stopwords_count}) else: st.warning(⚠️ 当前未配置停用词) with tab2: st.subheader(情感词配置) st.markdown(配置正面词汇和负面词汇用于情感分析) st.warning(⚠️ 未配置情感词将无法进行准确的情感分析) col1, col2 st.columns(2) with col1: st.markdown(#### 正面词汇) # 获取当前正面词汇 current_positive if user_positive_words in st.session_state: current_positive st.session_state.user_positive_words positive_input st.text_area( 正面词汇列表每行一个词, valuecurrent_positive, height200, help输入正面情感词汇每行一个。例如促进、推动、发展、提升、加强、完善、优化、创新、支持、鼓励 ) # 正面词汇文件上传 positive_file st.file_uploader(上传正面词汇文件, type[txt], keypositive_upload) if positive_file is not None: positive_content positive_file.getvalue().decode(utf-8) if st.button( 使用上传的正面词汇, keyuse_positive): st.session_state.user_positive_words positive_content st.success(✅ 已使用上传的正面词汇文件) st.rerun() with col2: st.markdown(#### 负面词汇) # 获取当前负面词汇 current_negative if user_negative_words in st.session_state: current_negative st.session_state.user_negative_words negative_input st.text_area( 负面词汇列表每行一个词, valuecurrent_negative, height200, help输入负面情感词汇每行一个。例如限制、禁止、约束、处罚、罚款、停止、取消、减少、降低、防止 ) # 负面词汇文件上传 negative_file st.file_uploader(上传负面词汇文件, type[txt], keynegative_upload) if negative_file is not None: negative_content negative_file.getvalue().decode(utf-8) if st.button( 使用上传的负面词汇, keyuse_negative): st.session_state.user_negative_words negative_content st.success(✅ 已使用上传的负面词汇文件) st.rerun() # 保存按钮 col1, col2 st.columns(2) with col1: if st.button( 保存情感词配置, typeprimary): st.session_state.user_positive_words positive_input st.session_state.user_negative_words negative_input # 保存到文件 positive_list [word.strip() for word in positive_input.split(\n) if word.strip()] negative_list [word.strip() for word in negative_input.split(\n) if word.strip()] st.session_state.config_manager.save_user_config(positive_words, positive_list) st.session_state.config_manager.save_user_config(negative_words, negative_list) st.success(✅ 情感词配置已保存) with col2: if st.button(️ 清空情感词): st.session_state.user_positive_words st.session_state.user_negative_words st.success(✅ 已清空情感词) st.rerun() # 显示统计信息 col1, col2 st.columns(2) with col1: if positive_input: positive_count len([word for word in positive_input.split(\n) if word.strip()]) st.info(f正面词汇数量: {positive_count}) else: st.warning(⚠️ 未配置正面词汇) with col2: if negative_input: negative_count len([word for word in negative_input.split(\n) if word.strip()]) st.info(f负面词汇数量: {negative_count}) else: st.warning(⚠️ 未配置负面词汇) with tab3: st.subheader(特征关键词库配置) st.markdown(配置政策特征关键词库用于自动评分。JSON格式{\特征类别\: [\关键词1\, \关键词2\, ...]}) st.warning(⚠️ 未配置特征关键词库将无法进行自动评分) # 获取当前关键词库 current_keywords if user_feature_keywords in st.session_state: current_keywords st.session_state.user_feature_keywords else: # 提供示例 current_keywords json.dumps({ 预测性: [预测, 预计, 预期, 展望, 规划], 监管性: [监管, 管理, 监督, 检查, 规范], 人才引进: [人才, 引进, 培养, 培训, 专家], 资金支持: [资金, 补贴, 资助, 融资, 贷款] }, ensure_asciiFalse, indent2) # 编辑关键词库 keywords_input st.text_area( 特征关键词库JSON格式, valuecurrent_keywords, height400, help输入JSON格式的特征关键词库 ) col1, col2, col3 st.columns(3) with col1: if st.button( 保存关键词库, typeprimary): if keywords_input: try: keywords_dict json.loads(keywords_input) if isinstance(keywords_dict, dict): st.session_state.user_feature_keywords keywords_input st.session_state.config_manager.save_user_config(feature_keywords, keywords_dict) st.success(✅ 特征关键词库已保存) else: st.error(关键词库必须是JSON字典格式) except json.JSONDecodeError as e: st.error(fJSON格式错误: {str(e)}) else: st.warning(请输入关键词库) with col2: # 文件上传 keywords_file st.file_uploader(上传关键词库文件, type[json, txt]) if keywords_file is not None: keywords_content keywords_file.getvalue().decode(utf-8) st.text_area(文件内容预览, keywords_content[:500], height150) if st.button( 使用上传文件): try: keywords_dict json.loads(keywords_content) if isinstance(keywords_dict, dict): st.session_state.user_feature_keywords keywords_content st.session_state.config_manager.save_user_config(feature_keywords, keywords_dict) st.success(✅ 已使用上传的关键词库文件) st.rerun() else: st.error(文件必须是JSON字典格式) except json.JSONDecodeError as e: st.error(fJSON格式错误: {str(e)}) with col3: if st.button(️ 清空关键词库): st.session_state.user_feature_keywords st.success(✅ 已清空关键词库) st.rerun() # 显示统计信息 if keywords_input: try: keywords_dict json.loads(keywords_input) if isinstance(keywords_dict, dict): category_count len(keywords_dict) total_keywords sum(len(keywords) for keywords in keywords_dict.values()) st.info(f特征类别数量: {category_count}, 总关键词数量: {total_keywords}) else: st.warning(⚠️ 关键词库格式不正确) except: st.warning(⚠️ 关键词库格式不正确) else: st.warning(⚠️ 未配置特征关键词库) with tab4: st.subheader(配置导入/导出) st.markdown(导入或导出所有配置) col1, col2 st.columns(2) with col1: st.markdown(### 导出配置) st.markdown(导出当前所有词汇库配置为ZIP文件) if st.button( 导出所有配置): try: zip_path st.session_state.config_manager.export_all_configs() with open(zip_path, rb) as f: zip_data f.read() st.download_button( label 下载配置ZIP文件, datazip_data, file_namepmc_configs.zip, mimeapplication/zip ) except Exception as e: st.error(f导出配置失败: {str(e)}) with col2: st.markdown(### 导入配置) st.markdown(从ZIP文件导入配置将覆盖当前配置) uploaded_zip st.file_uploader(选择配置ZIP文件, type[zip]) if uploaded_zip is not None: if st.button( 导入配置): try: configs st.session_state.config_manager.import_configs_from_zip(uploaded_zip) # 更新session state if stopwords in configs: st.session_state.user_stopwords \n.join(configs[stopwords]) if positive_words in configs: st.session_state.user_positive_words \n.join(configs[positive_words]) if negative_words in configs: st.session_state.user_negative_words \n.join(configs[negative_words]) if feature_keywords in configs: st.session_state.user_feature_keywords json.dumps(configs[feature_keywords], ensure_asciiFalse, indent2) if pmc_weights in configs: st.session_state.user_pmc_weights json.dumps(configs[pmc_weights], ensure_asciiFalse, indent2) if evaluation_criteria in configs: st.session_state.user_evaluation_criteria json.dumps(configs[evaluation_criteria], ensure_asciiFalse, indent2) st.success(✅ 配置导入成功) st.rerun() except Exception as e: st.error(f导入配置失败: {str(e)}) # PMC评估页面 elif page PMC评估: st.header(PMC指数评估) if not st.session_state.policies: st.info(暂无政策数据请先在政策管理中添加政策。) else: # 检查词汇配置 config_warnings [] if user_stopwords not in st.session_state or not st.session_state.user_stopwords: config_warnings.append(停用词) if user_feature_keywords not in st.session_state or not st.session_state.user_feature_keywords: config_warnings.append(特征关键词库) if config_warnings: st.warning(f⚠️ 以下词汇库未配置可能影响评估准确性{, .join(config_warnings)}。请先配置词汇库。) # 选择政策 policy_options {f{p[id]}: {p[name]}: p[id] for p in st.session_state.policies} selected_policy_key st.selectbox(选择要评估的政策, list(policy_options.keys())) selected_policy_id policy_options[selected_policy_key] selected_policy next(p for p in st.session_state.policies if p[id] selected_policy_id) st.subheader(f政策评估: {selected_policy[name]}) # 创建评估表单 with st.form(pmc_evaluation_form): st.write(请为政策的各个维度评分0-1分1分为最高:) # 动态生成评分滑块 pmc_data {} for primary_var, primary_name in st.session_state.pmc_model.primary_variables.items(): st.markdown(f### {primary_var}: {primary_name}) sec_vars st.session_state.pmc_model.secondary_variables.get(primary_var, []) cols st.columns(len(sec_vars)) pmc_data[primary_var] {} for i, sec_var in enumerate(sec_vars): with cols[i]: # 提取显示名称 display_name sec_var.replace(f{primary_var}_, ).replace(_, ) # 获取当前值 current_value 0.5 if selected_policy_id in st.session_state.pmc_results: suggestions st.session_state.pmc_results[selected_policy_id].get(pmc_suggestions, {}) if primary_var in suggestions and sec_var in suggestions[primary_var]: current_value suggestions[primary_var][sec_var] # 创建滑块 score st.slider( display_name, min_value0.0, max_value1.0, valuecurrent_value, step0.1, keyf{selected_policy_id}_{sec_var} ) pmc_data[primary_var][sec_var] score submitted st.form_submit_button(计算PMC指数) if submitted: # 计算PMC指数 pmc_score st.session_state.pmc_model.calculate_pmc_index(pmc_data) level, color st.session_state.pmc_model.evaluate_policy_level(pmc_score) variable_scores st.session_state.pmc_model.calculate_variable_scores(pmc_data) # 保存结果 st.session_state.pmc_results[selected_policy_id] { pmc_score: pmc_score, level: level, color: color, variable_scores: variable_scores, pmc_suggestions: pmc_data } st.success(fPMC指数计算完成: **{pmc_score:.2f}** ({level})) # 显示评估结果 if selected_policy_id in st.session_state.pmc_results: result st.session_state.pmc_results[selected_policy_id] col1, col2, col3 st.columns(3) with col1: st.metric(PMC指数, f{result[pmc_score]:.2f}) with col2: st.metric(评估等级, result[level]) with col3: # 显示颜色指示 st.markdown( fdiv stylebackground-color:{result[color]}; padding:10px; border-radius:5px; text-align:center; color:white;等级: {result[level]}/div, unsafe_allow_htmlTrue) # 显示详细得分 st.subheader(各维度得分详情) # 创建数据框 score_data [] for primary_var, score in result[variable_scores].items(): score_data.append({ 一级变量: f{primary_var}: {st.session_state.pmc_model.primary_variables[primary_var]}, 得分: score, 权重: st.session_state.pmc_model.weights.get(primary_var, 0), 加权得分: score * st.session_state.pmc_model.weights.get(primary_var, 0) * 9 }) score_df pd.DataFrame(score_data) st.dataframe(score_df, use_container_widthTrue) # 创建雷达图 st.subheader(维度得分雷达图) variables list(st.session_state.pmc_model.primary_variables.keys()) var_names [st.session_state.pmc_model.primary_variables[v] for v in variables] scores [result[variable_scores].get(v, 0) for v in variables] # 创建闭合数据 scores_closed scores [scores[0]] var_names_closed var_names [var_names[0]] fig go.Figure(data go.Scatterpolar( rscores_closed, thetavar_names_closed, filltoself, name变量得分, linedict(colorblue, width2), markerdict(size8, colorred) ) ) fig.update_layout( polardict( radialaxisdict( visibleTrue, range[0, 1] ) ), showlegendTrue, titlefPMC变量得分雷达图 - {selected_policy[name]}, height400 ) st.plotly_chart(fig, use_container_widthTrue) # 续写可视化分析页面 elif page 可视化分析: st.header(可视化分析) if not st.session_state.policies or not st.session_state.pmc_results: st.info(暂无评估数据请先在PMC评估中评估政策。) else: tab1, tab2, tab3 st.tabs([ PMC曲面图, 对比分析, 热力图]) with tab1: st.subheader(PMC曲面图) # 选择要可视化的政策 policy_options {f{p[id]}: {p[name]}: p[id] for p in st.session_state.policies if p[id] in st.session_state.pmc_results} if policy_options: selected_policy_key st.selectbox(选择政策, list(policy_options.keys()), keysurface_select) selected_policy_id policy_options[selected_policy_key] if selected_policy_id in st.session_state.pmc_results: result st.session_state.pmc_results[selected_policy_id] variable_scores result[variable_scores] # 准备曲面图数据 variables list(st.session_state.pmc_model.primary_variables.keys()) var_names [st.session_state.pmc_model.primary_variables[v] for v in variables] # 创建网格数据 theta np.linspace(0, 2 * np.pi, len(variables), endpointFalse) theta np.concatenate((theta, [theta[0]])) # 闭合多边形 # 获取得分并标准化 scores [variable_scores.get(v, 0) for v in variables] scores np.concatenate((scores, [scores[0]])) # 闭合多边形 # 创建极坐标图 fig go.Figure(data go.Scatterpolar( rscores, thetavar_names [var_names[0]], # 闭合 modelinesmarkers, filltoself, name变量得分, linedict(colorblue, width2), markerdict(size8, colorred) ) ) fig.update_layout( polardict( radialaxisdict( visibleTrue, range[0, 1] ) ), showlegendTrue, titlefPMC变量得分雷达图 - {selected_policy_key}, height500 ) st.plotly_chart(fig, use_container_widthTrue) # 创建3D曲面图 st.subheader(3D PMC曲面图) # 创建网格数据 x np.array([i for i in range(len(variables))]) y np.array([i for i in range(len(variables))]) X, Y np.meshgrid(x, y) # 创建Z矩阵这里使用简单的插值 Z np.zeros((len(variables), len(variables))) for i in range(len(variables)): for j in range(len(variables)): Z[i, j] (scores[i] scores[j]) / 2 fig_3d go.Figure(data[go.Surface(zZ, xx, yy, colorscaleViridis)]) fig_3d.update_layout( titlefPMC指数3D曲面图 - {selected_policy_key}, scenedict( xaxisdict(title变量索引, ticktextvar_names, tickvalslist(range(len(variables)))), yaxisdict(title变量索引, ticktextvar_names, tickvalslist(range(len(variables)))), zaxisdict(title得分, range[0, 1]) ), height600 ) st.plotly_chart(fig_3d, use_container_widthTrue) else: st.info(暂无评估数据可用于可视化。) with tab2: st.subheader(对比分析) if len(st.session_state.pmc_results) 2: # 创建对比表格 comparison_data [] for policy in st.session_state.policies: if policy[id] in st.session_state.pmc_results: result st.session_state.pmc_results[policy[id]] row { 政策名称: policy[name], 层级: policy[level], 地区: policy[region], PMC指数: result[pmc_score], 评估等级: result[level] } # 添加各维度得分 for var, score in result[variable_scores].items(): row[st.session_state.pmc_model.primary_variables[var]] f{score:.2f} comparison_data.append(row) if comparison_data: df_comparison pd.DataFrame(comparison_data) st.dataframe(df_comparison, use_container_widthTrue) # 创建柱状图对比 st.subheader(PMC指数对比图) fig go.Figure() policies df_comparison[政策名称].tolist() scores df_comparison[PMC指数].tolist() # 为不同等级设置颜色 colors [] for level in df_comparison[评估等级]: if level 优秀: colors.append(#4CAF50) elif level 良好: colors.append(#8BC34A) elif level 及格: colors.append(#FFC107) else: colors.append(#F44336) fig.add_trace(go.Bar( xpolicies, yscores, textscores, textpositionauto, marker_colorcolors )) fig.update_layout( titlePMC指数对比, xaxis_title政策, yaxis_titlePMC指数, yaxis_range[0, 9], height400 ) # 添加等级线 fig.add_hline(y8.0, line_dashdash, line_colorgreen, annotation_text优秀线) fig.add_hline(y7.0, line_dashdash, line_colororange, annotation_text良好线) fig.add_hline(y6.0, line_dashdash, line_colorred, annotation_text及格线) st.plotly_chart(fig, use_container_widthTrue) # 创建雷达图对比 st.subheader(维度得分雷达图对比) # 选择多个政策对比 policy_options {f{p[id]}: {p[name]}: p[id] for p in st.session_state.policies if p[id] in st.session_state.pmc_results} if len(policy_options) 2: selected_policies st.multiselect( 选择要对比的政策至少2个, list(policy_options.keys()), defaultlist(policy_options.keys())[:2] if len(policy_options) 2 else [] ) if len(selected_policies) 2: # 创建雷达图 fig_radar go.Figure() for policy_name in selected_policies: policy_id policy_options[policy_name] if policy_id in st.session_state.pmc_results: result st.session_state.pmc_results[policy_id] variable_scores result[variable_scores] variables list(st.session_state.pmc_model.primary_variables.keys()) var_names [st.session_state.pmc_model.primary_variables[v] for v in variables] scores [variable_scores.get(v, 0) for v in variables] scores scores [scores[0]] # 闭合 fig_radar.add_trace(go.Scatterpolar( rscores, thetavar_names [var_names[0]], namef{policy_name} ({result[pmc_score]:.2f}), filltoself )) fig_radar.update_layout( polardict( radialaxisdict( visibleTrue, range[0, 1] ) ), showlegendTrue, title政策对比雷达图, height500 ) st.plotly_chart(fig_radar, use_container_widthTrue) else: st.warning(请至少选择2个政策进行对比。) else: st.info(需要至少2个已评估的政策才能进行对比分析。) with tab3: st.subheader(热力图分析) if len(st.session_state.pmc_results) 2: # 准备热力图数据 policy_names [] variables list(st.session_state.pmc_model.primary_variables.keys()) var_names [st.session_state.pmc_model.primary_variables[v] for v in variables] scores_matrix [] for policy in st.session_state.policies: if policy[id] in st.session_state.pmc_results: policy_names.append(policy[name]) result st.session_state.pmc_results[policy[id]] variable_scores result[variable_scores] scores [variable_scores.get(v, 0) for v in variables] scores_matrix.append(scores) if scores_matrix: scores_matrix np.array(scores_matrix) # 创建热力图 fig go.Figure(datago.Heatmap( zscores_matrix, xvar_names, ypolicy_names, colorscaleRdYlGn, zmin0, zmax1, textnp.round(scores_matrix, 2), texttemplate%{text:.2f}, textfont{size: 10}, hovertemplate政策: %{y}br变量: %{x}br得分: %{z:.2f}extra/extra )) fig.update_layout( title政策变量得分热力图, xaxisdict( title评估变量, tickangle-45 ), yaxisdict( title政策名称 ), height400 len(policy_names) * 20, width800 ) st.plotly_chart(fig, use_container_widthTrue) # 创建相关矩阵 st.subheader(变量相关矩阵) # 计算相关系数矩阵 corr_matrix np.corrcoef(scores_matrix, rowvarFalse) fig_corr go.Figure(datago.Heatmap( zcorr_matrix, xvar_names, yvar_names, colorscaleRdBu, zmid0, zmin-1, zmax1, textnp.round(corr_matrix, 2), texttemplate%{text:.2f}, textfont{size: 10}, hovertemplateX: %{x}brY: %{y}br相关系数: %{z:.3f}extra/extra )) fig_corr.update_layout( title变量相关系数矩阵, xaxisdict( tickangle-45 ), height500, width600 ) st.plotly_chart(fig_corr, use_container_widthTrue) else: st.info(暂无数据可用于热力图分析。) else: st.info(需要至少2个已评估的政策才能进行热力图分析。) # 高级配置页面 elif page ⚙️ 高级配置: st.header(高级配置) st.markdown(配置PMC模型权重和评估标准) tab1, tab2, tab3 st.tabs([⚖️ PMC权重配置, 评估标准配置, 系统设置]) with tab1: st.subheader(PMC权重配置) st.markdown(配置一级变量权重和二级变量权重) col1, col2 st.columns(2) with col1: st.markdown(#### 一级变量权重) st.markdown(配置9个一级变量的权重总和应为1.0) # 获取当前权重 current_weights if user_pmc_weights in st.session_state and st.session_state.user_pmc_weights: current_weights st.session_state.user_pmc_weights else: # 使用等权重作为示例 num_vars len(st.session_state.pmc_model.primary_variables) equal_weight 1.0 / num_vars weights_dict {var: round(equal_weight, 3) for var in st.session_state.pmc_model.primary_variables.keys()} current_weights json.dumps(weights_dict, ensure_asciiFalse, indent2) weights_input st.text_area( 一级变量权重JSON格式, valuecurrent_weights, height300, help例如{X1: 0.1, X2: 0.1, X3: 0.12, X4: 0.12, X5: 0.12, X6: 0.12, X7: 0.1, X8: 0.1, X9: 0.12} ) if st.button( 保存权重配置, typeprimary): if weights_input: try: weights_dict json.loads(weights_input) if isinstance(weights_dict, dict): # 验证权重总和 total sum(weights_dict.values()) if abs(total - 1.0) 0.01: # 允许小误差 st.session_state.user_pmc_weights weights_input st.session_state.pmc_model.update_weights(weights_input) st.success(✅ 权重配置已保存) else: st.error(f权重总和应为1.0当前为{total:.2f}请调整权重。) else: st.error(权重配置必须是JSON字典格式) except json.JSONDecodeError as e: st.error(fJSON格式错误: {str(e)}) else: st.warning(请输入权重配置) with col2: st.markdown(#### 二级变量权重) st.markdown(系统自动计算二级变量权重通常无需手动配置) # 显示当前二级权重 secondary_weights st.session_state.pmc_model.secondary_weights if secondary_weights: st.markdown(当前二级权重配置) for var, weights in secondary_weights.items(): var_name st.session_state.pmc_model.primary_variables.get(var, var) sec_vars st.session_state.pmc_model.secondary_variables.get(var, []) st.markdown(f**{var_name} ({var})**:) for i, (sec_var, weight) in enumerate(zip(sec_vars, weights)): display_name sec_var.replace(f{var}_, ).replace(_, ) st.markdown(f- {display_name}: {weight:.2f}) # 权重验证 st.markdown(---) st.markdown(#### 权重验证) if st.button( 验证权重配置): if user_pmc_weights in st.session_state and st.session_state.user_pmc_weights: try: weights_dict json.loads(st.session_state.user_pmc_weights) total sum(weights_dict.values()) if abs(total - 1.0) 0.01: st.success(f✅ 权重配置有效总和为: {total:.2f}) else: st.error(f❌ 权重配置无效总和为: {total:.2f}应为1.0) except: st.error(❌ 权重配置格式错误) else: st.warning(⚠️ 未配置权重) # 权重文件上传 st.markdown(---) st.markdown(#### 权重文件上传) weights_file st.file_uploader(上传权重配置文件, type[json, txt]) if weights_file is not None: weights_content weights_file.getvalue().decode(utf-8) st.text_area(文件内容预览, weights_content[:500], height150) if st.button( 使用上传的权重文件): try: weights_dict json.loads(weights_content) if isinstance(weights_dict, dict): st.session_state.user_pmc_weights weights_content st.session_state.pmc_model.update_weights(weights_content) st.success(✅ 已使用上传的权重配置文件) st.rerun() else: st.error(文件必须是JSON字典格式) except json.JSONDecodeError as e: st.error(fJSON格式错误: {str(e)}) with tab2: st.subheader(评估标准配置) st.markdown(配置PMC指数的评估等级标准) # 获取当前评估标准 current_criteria if user_evaluation_criteria in st.session_state and st.session_state.user_evaluation_criteria: current_criteria st.session_state.user_evaluation_criteria else: # 使用默认评估标准 current_criteria json.dumps({ levels: { 优秀: {min: 8.0, max: 9.0, color: #4CAF50}, 良好: {min: 7.0, max: 8.0, color: #8BC34A}, 及格: {min: 6.0, max: 7.0, color: #FFC107}, 较差: {min: 0.0, max: 6.0, color: #F44336} } }, ensure_asciiFalse, indent2) criteria_input st.text_area( 评估标准配置JSON格式, valuecurrent_criteria, height300, help配置评估等级每个等级包含min最小值包含、max最大值不包含和color颜色 ) col1, col2 st.columns(2) with col1: if st.button( 保存评估标准, typeprimary): if criteria_input: try: criteria_dict json.loads(criteria_input) if isinstance(criteria_dict, dict) and levels in criteria_dict: st.session_state.user_evaluation_criteria criteria_input st.session_state.pmc_model.update_evaluation_criteria(criteria_input) st.success(✅ 评估标准已保存) else: st.error(评估标准必须包含levels字段) except json.JSONDecodeError as e: st.error(fJSON格式错误: {str(e)}) else: st.warning(请输入评估标准) with col2: if st.button(️ 恢复默认评估标准): default_criteria { levels: { 优秀: {min: 8.0, max: 9.0, color: #4CAF50}, 良好: {min: 7.0, max: 8.0, color: #8BC34A}, 及格: {min: 6.0, max: 7.0, color: #FFC107}, 较差: {min: 0.0, max: 6.0, color: #F44336} } } st.session_state.user_evaluation_criteria json.dumps(default_criteria, ensure_asciiFalse, indent2) st.success(✅ 已恢复默认评估标准) st.rerun() # 显示当前评估标准 st.markdown(---) st.markdown(#### 当前评估标准) try: criteria_dict json.loads(criteria_input) if levels in criteria_dict: levels criteria_dict[levels] for level_name, level_info in levels.items(): col1, col2, col3 st.columns(3) with col1: st.metric(等级, level_name) with col2: st.metric(分数范围, f{level_info.get(min, 0)}-{level_info.get(max, 0)}) with col3: color level_info.get(color, #000000) st.markdown( fdiv stylebackground-color:{color}; padding:10px; border-radius:5px; text-align:center; color:white;颜色/div, unsafe_allow_htmlTrue) except: st.warning(⚠️ 评估标准格式不正确) with tab3: st.subheader(系统设置) st.markdown(#### 数据管理) col1, col2 st.columns(2) with col1: if st.button( 保存当前会话): # 保存当前所有数据 session_data { policies: st.session_state.policies, pmc_results: st.session_state.pmc_results, analysis_history: st.session_state.analysis_history, user_stopwords: st.session_state.get(user_stopwords, ), user_positive_words: st.session_state.get(user_positive_words, ), user_negative_words: st.session_state.get(user_negative_words, ), user_feature_keywords: st.session_state.get(user_feature_keywords, ), user_pmc_weights: st.session_state.get(user_pmc_weights, ), user_evaluation_criteria: st.session_state.get(user_evaluation_criteria, ), saved_at: datetime.now().isoformat() } # 转换为JSON session_json json.dumps(session_data, ensure_asciiFalse, indent2) # 提供下载 st.download_button( label 下载会话数据, datasession_json, file_namefpmc_session_{datetime.now().strftime(%Y%m%d_%H%M%S)}.json, mimeapplication/json ) with col2: if st.button( 加载会话数据): uploaded_session st.file_uploader(选择会话文件, type[json]) if uploaded_session is not None: try: session_data json.load(uploaded_session) # 恢复会话数据 if policies in session_data: st.session_state.policies session_data[policies] if pmc_results in session_data: st.session_state.pmc_results session_data[pmc_results] if analysis_history in session_data: st.session_state.analysis_history session_data[analysis_history] # 恢复配置 if user_stopwords in session_data: st.session_state.user_stopwords session_data[user_stopwords] if user_positive_words in session_data: st.session_state.user_positive_words session_data[user_positive_words] if user_negative_words in session_data: st.session_state.user_negative_words session_data[user_negative_words] if user_feature_keywords in session_data: st.session_state.user_feature_keywords session_data[user_feature_keywords] if user_pmc_weights in session_data: st.session_state.user_pmc_weights session_data[user_pmc_weights] if user_evaluation_criteria in session_data: st.session_state.user_evaluation_criteria session_data[user_evaluation_criteria] st.success(✅ 会话数据加载成功) st.rerun() except Exception as e: st.error(f加载会话数据失败: {str(e)}) st.markdown(---) st.markdown(#### 系统信息) col1, col2, col3 st.columns(3) with col1: st.metric(Python版本, f{sys.version.split()[0]}) with col2: st.metric(Streamlit版本, f{st.__version__}) with col3: st.metric(Pandas版本, f{pd.__version__}) st.markdown(---) st.markdown(#### 系统重置) if st.button( 重置所有数据, typesecondary): if st.checkbox(确认重置所有数据此操作不可撤销。): st.session_state.policies [] st.session_state.pmc_results {} st.session_state.analysis_history [] st.session_state.user_stopwords st.session_state.user_positive_words st.session_state.user_negative_words st.session_state.user_feature_keywords st.session_state.user_pmc_weights st.session_state.user_evaluation_criteria st.success(✅ 所有数据已重置) st.rerun() # 页脚 st.markdown(---) st.caption(PMC政策文本量化评估系统 v1.0 | 完全可配置版本 | 所有词汇表由用户自行配置)