AI Forecast Improve

ai/EstimateMarketTrendV20.py

@@ -20,12 +20,17 @@ toForecast = False  # False means test, True means forecast
 syncData = False  # 开启会同步数据库指数及基金数据
 uploadData = False  # 开启会上传预测结果
 doReport = True  # 开启会生成Excel报告
+# JIFU_SPX_OPEPS_CURRQ_TTM
+# JIFU_SPX_OPEPS_YMAE1
+# JIFU_SPX_OPEPS_YMAE2
+# JIFU_SPX_OPEPS_YMAE3
+eps_eco = 'JIFU_SPX_OPEPS_CURRQ_TTM'  # 选取的eps指标名称
 
 # 待预测指数
 PREDICT_LIST = [67]
 # PREDICT_LIST = [67, 121, 122, 123, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163,
 #                 164, 165, 166, 167, 168, 169, 170, 171, 174, 175, 177, 178]
-eco = [65, 66, 74, 134, 191]
+eco = [65, 66, 74, 134, 191, 192, 193, 194]
 index = [67, 68, 69, 70, 71, 72, 73, 75, 76, 77, 105, 106, 116, 117, 138, 139, 142, 143, 140, 141, 144, 145, 146]
 # fund = [121, 122, 123, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160, 161, 162, 163, 164, 165,
 #         166, 167, 168, 169, 170, 171, 174, 175, 177, 178]

ai/reporter.py

 import datetime
 import json
+import math
 import random
 
 import pandas as pd
@@ -58,11 +59,13 @@ def do_reporter2(records=None, excel_name=None):
         datas.append(data)
     RMSE = get_RMSE(datas)
     RefRMSE = get_RefRMSE(datas)
+    RevisedReferenceRMSE = get_RevisedReferenceRMSE(datas)
     symbol_index_dict = {symbol: index for index, symbol in enumerate(symbols)}
     sorted_data = sorted(datas, key=lambda x: symbol_index_dict[x['Ticker'].split(' ')[0]])
     pf = pd.DataFrame(sorted_data)
     pf['RMSE'] = RMSE
     pf['RefRMSE'] = RefRMSE
+    pf['RevisedReferenceRMSE'] = RevisedReferenceRMSE
     pf.to_excel(excel_name if excel_name else "Forcast_Report_chu.xlsx", index=False)
 
 
@@ -98,6 +101,36 @@ def get_RefRMSE(datas):
          datas]) / len(datas)) ** 0.5
 
 
+def get_RevisedReferenceRMSE(datas):
+    """
+    Revised Reference RMSE
+    @param data: 预测值
+    @return:
+    """
+    datas = [data for data in datas if data.get('real outcome label') is not None]
+    tags = {v: k for k, v in LABEL_TAG.items()}
+
+    # Initialize the sum of squared differences
+    total_sum = 0
+
+    # Iterate over each data point
+    for data in datas:
+        # Get the true label
+        true_label = tags.get(data['real outcome label'])
+
+        # Sum up the squared differences for all possible deviations (-2 to 2)
+        for deviation in range(-2, 3):
+            total_sum += (true_label - deviation) ** 2
+
+    # Calculate the average by dividing by 5N
+    average = total_sum / (5 * len(datas))
+
+    # Take the square root to get the final Rev. Ref. RMSE
+    revised_ref_rmse = math.sqrt(average)
+
+    return revised_ref_rmse
+
+
 def is_right(id, type, start, predict):
     predict_term = 21
     navs = []

basic/sp500.py

@@ -112,7 +112,7 @@ def save_sp500():
         wb.close()
 
 
-def sync_sp500(day):
+def sync_sp500(ticker, day):
     file = Path(__file__).parent / 'resources/sp-500-eps-est_USA20241014.xlsx'
     if day:
         files = list_files_sorted_by_name(Path(__file__).parent / 'resources', day)
@@ -127,6 +127,7 @@ def sync_sp500(day):
     actuals = "ACTUALS"
     actuals_row = 0
     datas = []
+    estimates_datas = []
     # 遍历A列
     for row in range(100, 300):
         cell_value = ws[f'A{row}'].value
@@ -160,6 +161,13 @@ def sync_sp500(day):
                         'eps': ws[f'C{i}'].value,
                         'releaseDate': datetime.strptime(str(file)[-13:-5], "%Y%m%d")}
             datas.append(data)
+        else:
+            # 未来数据作为备用
+            data = {'date': date_value,
+                    'eps': ws[f'C{i}'].value}
+            data["releaseDate"] = data['date'] + timedelta(days=1)
+            data["date"] = data['releaseDate']
+            estimates_datas.append(data)
     for i in range(actuals_row + 1, ws.max_row):
         if ws[f'A{i}'].value is None:
             break
@@ -171,8 +179,22 @@ def sync_sp500(day):
         datas.append(data)
     wb.close()
     datas = pd.DataFrame(datas[::-1])
-    datas['eps'] = datas['eps'].shift(-1)
-    datas['close'] = datas['eps'].rolling(window=4).sum().round(2)
+    if ticker == 'JIFU SPX OPEPS CURRQ TTM':
+        datas['eps'] = datas['eps'].shift(-1)
+        datas['close'] = datas['eps'].rolling(window=4).sum().round(2)
+    elif ticker == 'JIFU SPX OPEPS YMAE1':
+        datas = datas.append(estimates_datas[:-2:-1], ignore_index=True)
+        datas['close'] = datas['eps'].rolling(window=4).sum().round(2)
+        datas['close'] = datas['close'].shift(-2)
+    elif ticker == 'JIFU SPX OPEPS YMAE2':
+        datas = datas.append(estimates_datas[:-3:-1], ignore_index=True)
+        datas['close'] = datas['eps'].rolling(window=4).sum().round(2)
+        datas['close'] = datas['close'].shift(-3)
+    elif ticker == 'JIFU SPX OPEPS YMAE3':
+        datas = datas.append(estimates_datas[:-4:-1], ignore_index=True)
+        datas['close'] = datas['eps'].rolling(window=4).sum().round(2)
+        datas['close'] = datas['close'].shift(-4)
+    datas = datas[datas['date'] <= report_day]
     datas.ffill(inplace=True)
     datas.dropna(inplace=True)
     return datas.to_dict(orient="records")[-1::] if day else datas.to_dict(orient="records")
@@ -192,4 +214,4 @@ def send_sp500(content, attach_paths):
 if __name__ == '__main__':
     # print(list_files_sorted_by_name(Path(__file__).parent / 'resources'))
     # save_sp500()
-    sync_sp500(day=None)
+    sync_sp500(ticker='JIFU SPX OPEPS YMAE3', day=None)

basic/sync.py

@@ -248,7 +248,7 @@ class EcoSync(EcoSync):
             if datum.get("source") == "calculating":
                 logger.debug(f'start sync ticker[{datum["bloombergTicker"]}]')
                 while True:
-                    datas = sync_sp500(self.datum_start_date(datum['id']))
+                    datas = sync_sp500(datum.get("bloombergTicker"), self.datum_start_date(datum['id']))
                     if datas:
                         self.store_date(datum['id'], datas)
                     else: