refactor: 清理测试文件和2019相关代码

- 删除测试脚本: test_2026_classification.py, test_lagllama.py, test_timesfm.py, test_ttm.py
- 删除2019相关文件: analyze_2019_*, apply_2019_*, dlinear_2019_*, three_models_2019_*
- 添加集成模型相关代码和分析脚本
- 添加N-BEATS模型和CatBoost分类模型

results/summary_report_2026_01.json

@@ -0,0 +1,58 @@
+{
+  "测试期间": {
+    "开始日期": "2026-01-01 00:00:00",
+    "结束日期": "2026-01-30 00:00:00",
+    "总天数": 22,
+    "总样本数": 495
+  },
+  "单模型性能": {
+    "DLinear": {
+      "准确率": "53.33%",
+      "预测上涨比例": "47.1%"
+    },
+    "N-BEATS": {
+      "准确率": "49.90%",
+      "预测上涨比例": "73.5%"
+    },
+    "TTM-R2": {
+      "准确率": "51.31%",
+      "预测上涨比例": "23.6%"
+    }
+  },
+  "集成策略性能": {
+    "双模型交集": {
+      "准确率": "53.39%",
+      "覆盖率": "47.68%",
+      "有效预测数": 236
+    },
+    "三模型投票": {
+      "准确率": "52.93%",
+      "覆盖率": "100.00%",
+      "有效预测数": 495
+    }
+  },
+  "DLinear概率分布": {
+    "0.4-0.5区间": "262个 (52.9%)",
+    "0.5-0.6区间": "229个 (46.3%)",
+    "0.6-0.7区间": "4个 (0.8%)",
+    "平均概率": "0.501"
+  },
+  "最佳/最差交易日": {
+    "双模型最佳日": {
+      "日期": "2026-01-11",
+      "准确率": "83.3%"
+    },
+    "双模型最差日": {
+      "日期": "2026-01-23",
+      "准确率": "28.6%"
+    },
+    "三模型最佳日": {
+      "日期": "2026-01-28",
+      "准确率": "80.0%"
+    },
+    "三模型最差日": {
+      "日期": "2026-01-23",
+      "准确率": "29.2%"
+    }
+  }
+}