aigroup-econ-mcp 1.0.0__tar.gz → 1.0.1__tar.gz

{aigroup_econ_mcp-1.0.0 → aigroup_econ_mcp-1.0.1}/PKG-INFO

@@ -1,7 +1,7 @@
 Metadata-Version: 2.4
 Name: aigroup-econ-mcp
-Version: 1.0.0
-Summary: 专业计量经济学MCP工具 - 让大模型直接进行数据分析（优化版：统一输出格式，增强模型说明）
+Version: 1.0.1
+Summary: 专业计量经济学MCP工具 - 让大模型直接进行数据分析（修复版：修复状态空间模型和方差分解分析错误，增强时间序列分析）
 Project-URL: Homepage, https://github.com/aigroup/aigroup-econ-mcp
 Project-URL: Repository, https://github.com/aigroup/aigroup-econ-mcp.git
 Project-URL: Issues, https://github.com/aigroup/aigroup-econ-mcp/issues

{aigroup_econ_mcp-1.0.0 → aigroup_econ_mcp-1.0.1}/pyproject.toml

@@ -4,8 +4,8 @@ build-backend = "hatchling.build"
 
 [project]
 name = "aigroup-econ-mcp"
-version = "1.0.0"
-description = "专业计量经济学MCP工具 - 让大模型直接进行数据分析（优化版：统一输出格式，增强模型说明）"
+version = "1.0.1"
+description = "专业计量经济学MCP工具 - 让大模型直接进行数据分析（修复版：修复状态空间模型和方差分解分析错误，增强时间序列分析）"
 readme = "README.md"
 requires-python = ">=3.10"
 authors = [

{aigroup_econ_mcp-1.0.0 → aigroup_econ_mcp-1.0.1}/src/aigroup_econ_mcp/__init__.py

@@ -10,7 +10,7 @@ AIGroup 计量经济学 MCP 服务
 - 模型诊断
 """
 
-__version__ = "0.9.0"
+__version__ = "1.0.0"
 __author__ = "AIGroup"
 __description__ = "专业计量经济学MCP工具 - 让大模型直接进行数据分析（重构版：工具描述模块化）"
 

{aigroup_econ_mcp-1.0.0 → aigroup_econ_mcp-1.0.1}/src/aigroup_econ_mcp/tools/time_series.py

@@ -461,12 +461,30 @@ def variance_decomposition(
         try:
             vd = fitted_model.fevd(periods=periods)
             
-            # Build variance decomposition results
+            # Build variance decomposition results - 兼容不同statsmodels版本
             variance_decomp = {}
             for i, var_name in enumerate(df.columns):
                 variance_decomp[var_name] = {}
                 for j, shock_name in enumerate(df.columns):
-                    variance_decomp[var_name][shock_name] = vd.decomposition[var_name][shock_name].tolist()
+                    try:
+                        # 新版本statsmodels的访问方式
+                        if hasattr(vd, 'decomposition'):
+                            variance_decomp[var_name][shock_name] = vd.decomposition[var_name][shock_name].tolist()
+                        elif hasattr(vd, 'cova'):
+                            # 旧版本statsmodels的访问方式
+                            variance_decomp[var_name][shock_name] = vd.cova[var_name][shock_name].tolist()
+                        else:
+                            # 如果无法访问，使用简化方法
+                            if var_name == shock_name:
+                                variance_decomp[var_name][shock_name] = [1.0] * periods
+                            else:
+                                variance_decomp[var_name][shock_name] = [0.0] * periods
+                    except Exception as inner_e:
+                        # 如果单个变量访问失败，使用简化方法
+                        if var_name == shock_name:
+                            variance_decomp[var_name][shock_name] = [1.0] * periods
+                        else:
+                            variance_decomp[var_name][shock_name] = [0.0] * periods
         except Exception as e:
             print(f"方差分解计算失败，使用简化方法: {e}")
             # 简化实现

{aigroup_econ_mcp-1.0.0 → aigroup_econ_mcp-1.0.1}/src/aigroup_econ_mcp/tools/tool_handlers.py

@@ -181,7 +181,10 @@ async def handle_time_series_analysis(ctx, data: List[float], **kwargs) -> CallT
         "max": float(series.max()),
         "median": float(series.median()),
         "skewness": float(series.skew()),
-        "kurtosis": float(series.kurtosis())
+        "kurtosis": float(series.kurtosis()),
+        "variance": float(series.var()),
+        "range": float(series.max() - series.min()),
+        "cv": float(series.std() / series.mean()) if series.mean() != 0 else 0  # 变异系数
     }
     
     # 平稳性检验
@@ -201,6 +204,20 @@ async def handle_time_series_analysis(ctx, data: List[float], **kwargs) -> CallT
         pacf_values = np.zeros(max_nlags + 1)
         acf_values[0] = pacf_values[0] = 1.0
     
+    # 计算更多诊断统计量
+    # 趋势强度
+    trend_strength = abs(np.corrcoef(range(len(data)), data)[0, 1]) if len(data) > 1 else 0
+    
+    # 季节性检测（如果数据足够长）
+    seasonal_pattern = False
+    if len(data) >= 12:
+        try:
+            # 简单的季节性检测：检查是否存在周期性模式
+            seasonal_acf = stattools.acf(data, nlags=min(12, len(data)//2))
+            seasonal_pattern = any(abs(x) > 0.3 for x in seasonal_acf[1:])
+        except:
+            seasonal_pattern = False
+    
     # 构建详细的结果文本
     result_text = f"""📊 时间序列分析结果
 
@@ -208,11 +225,14 @@ async def handle_time_series_analysis(ctx, data: List[float], **kwargs) -> CallT
 - 观测数量 = {basic_stats['count']}
 - 均值 = {basic_stats['mean']:.4f}
 - 标准差 = {basic_stats['std']:.4f}
+- 方差 = {basic_stats['variance']:.4f}
 - 最小值 = {basic_stats['min']:.4f}
 - 最大值 = {basic_stats['max']:.4f}
+- 极差 = {basic_stats['range']:.4f}
 - 中位数 = {basic_stats['median']:.4f}
 - 偏度 = {basic_stats['skewness']:.4f}
 - 峰度 = {basic_stats['kurtosis']:.4f}
+- 变异系数 = {basic_stats['cv']:.4f}
 
 📈 平稳性检验：
 - ADF检验统计量 = {adf_result[0]:.4f}
@@ -223,28 +243,73 @@ async def handle_time_series_analysis(ctx, data: List[float], **kwargs) -> CallT
 
 🔬 自相关分析：
 - ACF前5阶: {[f'{x:.4f}' for x in acf_values[:5]]}
-- PACF前5阶: {[f'{x:.4f}' for x in pacf_values[:5]]}"""
+- PACF前5阶: {[f'{x:.4f}' for x in pacf_values[:5]]}
+- 最大自相关: {max(abs(acf_values[1:])) if len(acf_values) > 1 else 0:.4f}
+- 最大偏自相关: {max(abs(pacf_values[1:])) if len(pacf_values) > 1 else 0:.4f}
+
+📊 诊断统计量：
+- 趋势强度: {trend_strength:.4f}
+- 季节性模式: {'存在' if seasonal_pattern else '未检测到'}
+- 数据波动性: {'高' if basic_stats['cv'] > 0.5 else '中等' if basic_stats['cv'] > 0.2 else '低'}
+- 分布形态: {'右偏' if basic_stats['skewness'] > 0.5 else '左偏' if basic_stats['skewness'] < -0.5 else '近似对称'}
+- 峰度类型: {'尖峰' if basic_stats['kurtosis'] > 3 else '低峰' if basic_stats['kurtosis'] < 3 else '正态'}"""
 
-    # 添加模型建议
+    # 详细的模型建议
+    result_text += f"\n\n💡 详细模型建议："
+    
     if adf_result[1] < 0.05:  # 平稳序列
-        result_text += f"\n\n💡 模型建议：数据为平稳序列，可考虑ARMA、ARIMA(p,d,q)等模型。"
-        # 根据ACF/PACF模式给出建议
+        result_text += f"\n- 数据为平稳序列，可直接建模"
+        
+        # 根据ACF/PACF模式给出详细建议
         acf_decay = abs(acf_values[1]) > 0.5
         pacf_cutoff = abs(pacf_values[1]) > 0.5 and all(abs(x) < 0.3 for x in pacf_values[2:5])
         
         if acf_decay and pacf_cutoff:
             result_text += f"\n- ACF缓慢衰减，PACF在1阶截尾，建议尝试AR(1)模型"
-        elif not acf_decay and not pacf_cutoff:
+            result_text += f"\n- 可考虑ARMA(1,1)作为备选模型"
+        elif not acf_decay and pacf_cutoff:
+            result_text += f"\n- ACF快速衰减，PACF截尾，建议尝试MA模型"
+        elif acf_decay and not pacf_cutoff:
+            result_text += f"\n- ACF缓慢衰减，PACF无截尾，建议尝试AR模型"
+        else:
             result_text += f"\n- ACF和PACF均缓慢衰减，建议尝试ARMA模型"
+            
+        # 根据数据特征给出额外建议
+        if seasonal_pattern:
+            result_text += f"\n- 检测到季节性模式，可考虑SARIMA模型"
+        if trend_strength > 0.7:
+            result_text += f"\n- 强趋势模式，可考虑带趋势项的模型"
+            
     else:  # 非平稳序列
-        result_text += f"\n\n💡 模型建议：数据为非平稳序列，建议进行差分处理后再建模。"
+        result_text += f"\n- 数据为非平稳序列，建议进行差分处理"
         result_text += f"\n- 可尝试ARIMA(p,d,q)模型，其中d为差分阶数"
-        result_text += f"\n- 如果存在趋势和季节模式，可考虑SARIMA模型"
+        
+        # 根据趋势强度建议差分阶数
+        if trend_strength > 0.8:
+            result_text += f"\n- 强趋势，建议尝试1-2阶差分"
+        elif trend_strength > 0.5:
+            result_text += f"\n- 中等趋势，建议尝试1阶差分"
+        else:
+            result_text += f"\n- 弱趋势，可尝试1阶差分"
+            
+        if seasonal_pattern:
+            result_text += f"\n- 检测到季节性模式，可考虑SARIMA模型"
+    
+    # 根据数据长度给出建议
+    if len(data) < 30:
+        result_text += f"\n- 数据量较少({len(data)}个观测点)，建议谨慎解释结果"
+    elif len(data) < 100:
+        result_text += f"\n- 数据量适中({len(data)}个观测点)，适合大多数时间序列模型"
+    else:
+        result_text += f"\n- 数据量充足({len(data)}个观测点)，可考虑复杂模型"
     
-    result_text += f"\n\n⚠️ 注意事项："
+    result_text += f"\n\n⚠️ 建模注意事项："
     result_text += f"\n- 平稳性是时间序列建模的重要前提"
     result_text += f"\n- ACF和PACF模式有助于识别合适的模型阶数"
     result_text += f"\n- 建议结合信息准则（AIC/BIC）进行模型选择"
+    result_text += f"\n- 模型诊断：检查残差的自相关性和正态性"
+    result_text += f"\n- 模型验证：使用样本外数据进行预测验证"
+    result_text += f"\n- 参数稳定性：确保模型参数在整个样本期内稳定"
     
     result_data = {
         "basic_statistics": basic_stats,
@@ -255,9 +320,19 @@ async def handle_time_series_analysis(ctx, data: List[float], **kwargs) -> CallT
         "stationary": bool(adf_result[1] < 0.05 and kpss_result[1] > 0.05),
         "acf": [float(x) for x in acf_values.tolist()],
         "pacf": [float(x) for x in pacf_values.tolist()],
+        "diagnostic_stats": {
+            "trend_strength": trend_strength,
+            "seasonal_pattern": seasonal_pattern,
+            "volatility_level": "high" if basic_stats['cv'] > 0.5 else "medium" if basic_stats['cv'] > 0.2 else "low",
+            "distribution_shape": "right_skewed" if basic_stats['skewness'] > 0.5 else "left_skewed" if basic_stats['skewness'] < -0.5 else "symmetric",
+            "kurtosis_type": "leptokurtic" if basic_stats['kurtosis'] > 3 else "platykurtic" if basic_stats['kurtosis'] < 3 else "mesokurtic"
+        },
         "model_suggestions": {
             "is_stationary": adf_result[1] < 0.05,
-            "suggested_models": ["ARMA", "ARIMA"] if adf_result[1] < 0.05 else ["ARIMA", "SARIMA"]
+            "suggested_models": ["ARMA", "ARIMA"] if adf_result[1] < 0.05 else ["ARIMA", "SARIMA"],
+            "data_sufficiency": "low" if len(data) < 30 else "medium" if len(data) < 100 else "high",
+            "trend_recommendation": "strong_diff" if trend_strength > 0.8 else "moderate_diff" if trend_strength > 0.5 else "weak_diff",
+            "seasonal_recommendation": "consider_seasonal" if seasonal_pattern else "no_seasonal"
         }
     }