feat(Config): feature extractors are enabled one-by-one with a bool, added previous model to model.fit() (#77)

* feat(Config): feature extractors are enabled one-by-one with a bool, added previous model to model.fit()

* fix(Sweep): removed unused `other_features` parameter that fails sweep

* feat(Config): using preset names for defining feature extractors again

* fix(Tests): fixed model stub classes

config.py

@@ -1,5 +1,6 @@
+from collections import defaultdict
 from utils.load_data import get_crypto_assets
-import feature_extractors.feature_extractor_presets as feature_extractor_presets
+from feature_extractors.feature_extractor_presets import presets
 from models.model_map import model_names_classification, model_names_regression
 
 def get_default_config() -> tuple[dict, dict, dict]:
@@ -18,8 +19,8 @@ def get_default_config() -> tuple[dict, dict, dict]:
         load_other_assets= False,
         log_returns= True,
         forecasting_horizon = 1,
-        own_features= feature_extractor_presets.date + feature_extractor_presets.level1,
-        other_features= [],
+        own_features = ['level_1', 'date_days'],
+        other_features = [],
         index_column= 'int',
         method= 'classification',
     )

feature_extractors/feature_extractor_presets.py

@@ -1,24 +1,49 @@
 from feature_extractors.feature_extractors import feature_lag, feature_mom, feature_ROC, feature_RSI, feature_STOD, feature_STOK, feature_vol, feature_day_of_month, feature_day_of_week, feature_month, feature_debug_future_lookahead
+from utils.typing import FeatureExtractorConfig
+from utils.helpers import flatten
 
-debug_future_lookahead = [('debug_future', feature_debug_future_lookahead, [1])]
+__presets = dict(
+    debug_future_lookahead = [('debug_future', feature_debug_future_lookahead, [1])],
+    single_mom = [('mom', feature_mom, [30])],
+    single_vol = [('vol', feature_vol, [30])],
+    mom = [('mom', feature_mom, [10, 20, 30, 60, 90])],
+    vol = [('vol', feature_vol, [10, 20, 30, 60])],
+    lags_up_to_5 = [('lag', feature_lag, [1,2,3,4,5])],
+    lags_up_to_10 = [('lag', feature_lag, [1,2,3,4,5,6,7,8,9,10])],
+    date_all = [
+        ('day_of_week', feature_day_of_week, [0]),
+        ('day_of_month', feature_day_of_month, [0]),
+        ('month', feature_month, [0])],
+    date_days = [
+        ('day_of_week', feature_day_of_week, [0]),
+        ('day_of_month', feature_day_of_month, [0]),
+    ],
+    roc = [('roc', feature_ROC, [10, 30])],
+    rsi = [('rsi', feature_ROC, [10, 30, 100])],
+    stod = [('stod', feature_STOD, [10, 30, 200])],
+    stok = [('stok', feature_STOK, [10, 30, 200])],
+)
 
-lags = [('lag', feature_lag, [1,2,3,4,5,6,7,8,9])]
+presets = __presets | dict(
+    level_1 = __presets["mom"] + __presets["vol"],
+    level_2 = __presets["mom"] + __presets["vol"] + __presets["roc"] + __presets["rsi"] + __presets["stod"] + __presets["stok"],
+)
 
-only_mom = [('mom', feature_mom, [30])]
+def preprocess_feature_extractors_config(data_dict: dict) -> dict:
+    keys = ['own_features', 'other_features']
+    for key in keys:
+        preset_names = data_dict[key]
+        data_dict[key] = flatten([presets[preset_name] for preset_name in preset_names])
+    return data_dict
 
-date = [
-    ('day_of_week', feature_day_of_week, [0]),
-    ('day_of_month', feature_day_of_month, [0]),
-    ('month', feature_month, [0])]
 
-level1 = [
-    ('mom', feature_mom, [10, 20, 30, 60, 90]),
-    ('vol', feature_vol, [10, 20, 30, 60]),
-]
+# Use this if ever we want to create an independent boolean for each featureextractor
+# def preprocess_feature_extractors_config(data_dict: dict) -> dict:
+#     prefixes = ['own_features', 'other_features']
+#     features_dict = dict()
+#     for prefix in prefixes:
+#         features_to_include = [key.replace(prefix + "_", "") for key, value in data_dict.items() if key.startswith(prefix) and value == True]
+#         features_dict[prefix] = flatten([presets[feature_name] for feature_name in features_to_include])
 
-level2 = level1 + [
-    ('roc', feature_ROC, [10, 30]),
-    ('rsi', feature_RSI, [10, 30, 100]),
-    ('stod', feature_STOD, [10, 30, 200]),
-    ('stok', feature_STOK, [10, 30, 200]),
-]
+#     data_dict = {k: v for k, v in data_dict.items() if not (k.startswith(prefixes[0]) or k.startswith(prefixes[1]))}
+#     return (data_dict | features_dict)

models/average.py

@@ -10,7 +10,7 @@ class StaticAverageModel(Model):
     data_scaling = 'unscaled'
     only_column = 'model_'
 
-    def fit(self, X, y):
+    def fit(self, X, y, prev_model):
         # This is a static model, it can' learn anything
         pass
 

models/base.py

@@ -11,7 +11,7 @@ class Model(ABC):
     only_column: Optional[str]
 
     @abstractmethod
-    def fit(self, X, y):
+    def fit(self, X, y, prev_model):
         pass
 
     @abstractmethod
@@ -32,7 +32,7 @@ class SKLearnModel(Model):
     def __init__(self, model):
         self.model = model
 
-    def fit(self, X, y):
+    def fit(self, X, y, prev_model):
         self.model.fit(X, y)
 
     def predict(self, X):

models/momentum.py

@@ -14,7 +14,7 @@ class StaticMomentumModel(Model):
         super().__init__()
         self.allow_short = allow_short
 
-    def fit(self, X, y):
+    def fit(self, X, y, prev_model):
         # This is a static model, it can' learn anything
         pass
 

models/naive.py

@@ -10,7 +10,7 @@ class StaticNaiveModel(Model):
     data_scaling = 'unscaled'
     only_column = None
 
-    def fit(self, X, y):
+    def fit(self, X, y, prev_model):
         # This is a static model, it can' learn anything
         pass
 

reporting/wandb.py

@@ -15,15 +15,16 @@ def launch_wandb(project_name:str, default_config:dict, sweep:bool=False):
         return wandb
     
     
-def seperate_configs(wandb: Optional[object], model_config:dict, training_config:dict, data_config:dict) -> tuple[dict,dict,dict]:
+def register_config_with_wandb(wandb: Optional[object], model_config:dict, training_config:dict, data_config:dict):
     config: dict = wandb.config
 
     if type(wandb) is not type(None):
-        for k in training_config: training_config[k] = config[k]
-        for k in model_config: model_config[k] = config[k]
-        # for k in data_config: data_config[k] = config[k]
-
-    return model_config, training_config, data_config
+        for k in training_config:
+            training_config[k] = config[k]
+        for k in model_config:
+            model_config[k] = config[k]
+        for k in data_config:
+            data_config[k] = config[k]
 
 
 def send_report_to_wandb(results: pd.DataFrame, wandb:Optional[object], project_name: str, model_name: str):

run_pipeline.py

@@ -1,8 +1,9 @@
 from utils.load_data import load_data
 import pandas as pd
 from training.training import run_single_asset_trainig
-from reporting.wandb import launch_wandb, send_report_to_wandb, seperate_configs
+from reporting.wandb import launch_wandb, send_report_to_wandb, register_config_with_wandb
 from models.model_map import map_model_name_to_function
+from feature_extractors.feature_extractor_presets import preprocess_feature_extractors_config
 from config import get_default_config, validate_config, get_model_name
 
 def setup_pipeline(project_name:str, with_wandb: bool, sweep: bool):
@@ -11,9 +12,10 @@ def setup_pipeline(project_name:str, with_wandb: bool, sweep: bool):
     wandb = None
     if with_wandb: 
         wandb = launch_wandb(project_name=project_name, default_config=dict(**model_config, **training_config, **data_config), sweep=sweep)
-        model_config, training_config, data_config = seperate_configs(wandb, model_config, training_config, data_config)
+        register_config_with_wandb(wandb, model_config, training_config, data_config)
 
     model_config = map_model_name_to_function(model_config, data_config['method'])
+    data_config = preprocess_feature_extractors_config(data_config)
     pipeline(project_name, wandb, sweep, model_config, training_config, data_config)  
     
 

sweep.yaml

@@ -15,10 +15,10 @@ parameters:
       values: [True, False]
       distribution: categorical
     sliding_window_size: 
-      values: [90, 130, 160, 180, 280, 380]
+      values: [180, 280, 380]
       distribution: categorical
     retrain_every: 
-      values: [14, 30, 60, 100]
+      values: [10, 20, 30]
       distribution: categorical
     scaler:
       values: ['minmax', 'normalize', 'minmax', 'standardize', 'none']
@@ -37,10 +37,6 @@ parameters:
     log_returns: 
       values: [True, False]
       distribution: categorical
-    own_features: 
-      value: []
-    other_features: 
-      value: []
     index_column: 
       value: 'int'
     level_1_models: 
@@ -49,4 +45,8 @@ parameters:
     level_2_models: 
       value: []
       distribution: constant
-
+    own_features:
+      values: [['only_mom', 'date_days'], [], ['level_1', 'date_days'], ['level_1', 'date_days', 'level_2']]
+      distribution: categorical
+    other_features:
+      value: []

tests/test_evaluation.py

@@ -42,7 +42,7 @@ class EvenOddStubModel(Model):
         super().__init__()
         self.window_length = window_length
 
-    def fit(self, X, y):
+    def fit(self, X, y, prev_model):
         assert len(X) == self.window_length
         for i in range(len(X)):
             assert y[i] == -1 if X[i][0] == 1 else 1

tests/test_walk_forward.py

@@ -40,7 +40,7 @@ class IncrementingStubModel(Model):
         super().__init__()
         self.window_length = window_length
 
-    def fit(self, X, y):
+    def fit(self, X, y, prev_model):
         assert len(X) == self.window_length
         for i in range(len(X)):
             assert X[i][0] + 1 == y[i]

training/walk_forward.py

@@ -58,7 +58,7 @@ def walk_forward_train_test(
                 X_slice = X_slice.to_numpy()
 
             current_model = model.clone()
-            current_model.fit(X_slice, y_slice.to_numpy())
+            current_model.fit(X_slice, y_slice.to_numpy(), models[index-1])
             iterations_before_retrain = retrain_every
         else:
             current_model = models[index-1]

utils/helpers.py

@@ -4,3 +4,6 @@ import numpy as np
 
 def get_first_valid_return_index(series: pd.Series) -> int:
     return np.where(np.logical_and(series != 0, np.logical_not(np.isnan(series))))[0][0]
+
+def flatten(list_of_lists: list) -> list:
+    return [item for sublist in list_of_lists for item in sublist]

utils/typing.py

@@ -4,4 +4,6 @@ import pandas as pd
 
 Period = int
 IsLogReturn = bool
-FeatureExtractor = Callable[[pd.DataFrame, Period, IsLogReturn], Union[pd.DataFrame, pd.Series]]
+FeatureExtractor = Callable[[pd.DataFrame, Period, IsLogReturn], Union[pd.DataFrame, pd.Series]]
+Name = str
+FeatureExtractorConfig = tuple[Name, FeatureExtractor, list[Period]]