feat(Transformations): added robust scaler (#223)

* feat(Transformations): added robust scaler

* fix(Config): set back default scaler to MinMax

config/presets.py

@@ -15,10 +15,10 @@ def get_default_config() -> RawConfig:
 
     return RawConfig(
         dimensionality_reduction_ratio=0.5,
-        n_features_to_select=30,
+        n_features_to_select=50,
         sliding_window_size=3800,
         retrain_every=1000,
-        scaler="minmax",  # 'normalize' 'minmax' 'standardize'
+        scaler="minmax",  # 'normalize' 'minmax' 'standardize' 'robust'
         assets=["fivemin_crypto"],
         target_asset="BTC_USD",
         other_assets=[],

config/types.py

@@ -3,7 +3,7 @@ from typing import Literal, Optional
 from labeling.types import EventFilter
 from models.base import Model
 from data_loader.types import DataCollection, DataSource
-from feature_extractors.types import FeatureExtractor, ScalerTypes
+from feature_extractors.types import FeatureExtractor
 from labeling.types import EventFilter, EventLabeller
 from sklearn.base import BaseEstimator
 from dataclasses import dataclass
@@ -15,7 +15,7 @@ class RawConfig(BaseModel):
     n_features_to_select: int
     sliding_window_size: int
     retrain_every: int
-    scaler: Literal["normalize", "minmax", "standardize"]
+    scaler: Literal["normalize", "minmax", "standardize", "robust"]
 
     assets: list[str]
     target_asset: str

data_loader/get_prices.py

@@ -16,10 +16,6 @@ def get_crypto_price_crypto_compare(
     return df
 
 
-ada = get_crypto_price_crypto_compare("ADA", "USD", 1500)
-ada
-
-
 def get_crypto_price_av(symbol: str, exchange: str, start_date=None) -> pd.DataFrame:
     api_url = f"https://www.alphavantage.co/query?function=DIGITAL_CURRENCY_DAILY&symbol={symbol}&market={exchange}&apikey={AV_API_KEY}"
     raw_df = requests.get(api_url).json()

feature_extractors/types.py

@@ -6,4 +6,3 @@ IsLogReturn = bool
 FeatureExtractor = Callable[[pd.DataFrame, Period], Union[pd.DataFrame, pd.Series]]
 Name = str
 FeatureExtractorConfig = tuple[Name, FeatureExtractor, list[Period]]
-ScalerTypes = Literal["normalize", "minmax", "standardize", "none"]

transformations/retrieve.py

@@ -3,9 +3,9 @@ from .pca import PCATransformation
 from .sklearn import SKLearnTransformation
 from typing import Literal, Optional
 from models.model_map import default_feature_selector_classification
-from sklearn.preprocessing import MinMaxScaler, Normalizer, StandardScaler
+from sklearn.preprocessing import MinMaxScaler, Normalizer, StandardScaler, RobustScaler
 
-ScalerTypes = Literal["normalize", "minmax", "standardize"]
+ScalerTypes = Literal["normalize", "minmax", "standardize", "robust"]
 
 
 def get_rfe(n_feature_to_select: int) -> Optional[RFETransformation]:
@@ -38,5 +38,9 @@ def get_scaler(type: ScalerTypes) -> SKLearnTransformation:
         return SKLearnTransformation(MinMaxScaler(feature_range=(-1, 1)))
     elif type == "standardize":
         return SKLearnTransformation(StandardScaler())
+    elif type == "robust":
+        return SKLearnTransformation(
+            RobustScaler(with_centering=False, quantile_range=(0.10, 0.90))
+        )
     else:
         raise Exception("Scaler type not supported")