1.1

- Added voting based b_s orders.
- Created config file (too much hassle otherwise)
- Created utils.py for shared utility functions (read_file, read_config).
- Small clean up
- Added way to find "best_parameters" on sample data (using param_test_script.py & param_test_inference.py

configuration.ini

@@ -17,10 +17,10 @@ delta_values = ['variable', 2, 3, 4, 5]
 months_out_of_sample = 36
 
 [actual_names]
-^GSPC = S&P 500
+^GSPC = S&P500
 ^DJI = Dow Jones Industrial Average
 ^GDAXI = DAX30
-^FCHI = CAC 40
+^FCHI = CAC40
 ^N225 = Nikkei 225
 brent oil = Brent
 natural gas = NG

inference.py

@@ -43,7 +43,7 @@ def vote_based_b_s(df, list_of_distance_metrics, list_of_stat_models):
     return df
 
 
-def main(stock_index):
+def main(stock_index, vote_based=False):
 
     next_returns, b_s_orders = utils.read_data(stock_index, True)
 
@@ -53,9 +53,10 @@ def main(stock_index):
     month_id = np.unique(b_s_orders["monthID"])
 
     # Create an equal vote based b_s order.
-    list_of_distance_metrics = ["dtw", "twed", "lcss"]
-    list_of_stat_models = ["knn"]
-    b_s_orders = vote_based_b_s(b_s_orders, list_of_distance_metrics, list_of_stat_models).reset_index()
+    if vote_based:
+        list_of_distance_metrics = ["dtw", "twed", "lcss"]
+        list_of_stat_models = ["knn"]
+        b_s_orders = vote_based_b_s(b_s_orders, list_of_distance_metrics, list_of_stat_models).reset_index()
 
     # Instead of multi-index using a "label" column.
     b_s_orders["Labels"] = b_s_orders["data_normalization"] + " " + b_s_orders["distance_model"] + " " + b_s_orders["stat_model"]
@@ -133,7 +134,6 @@ def main(stock_index):
 dict_indexes = utils.read_config("actual_names")
 
 if __name__ == '__main__':
-
     for instrument in instruments:
         main(instrument)
 

utils.py

@@ -37,7 +37,7 @@ def convert(x):
         try:
             return int(x)
         except ValueError:
-            pass
+            return x
     else:
         return x