Input models

openenergyid/__init__.py

@@ -1,6 +1,6 @@
 """Open Energy ID Python SDK."""
 
-__version__ = "0.1.6"
+__version__ = "0.1.7"
 
 from .enums import Granularity
 from .models import TimeSeries

openenergyid/mvlr/__init__.py

@@ -1,12 +1,19 @@
 """Multi-variable linear regression (MVLR) module."""
 
-from .mvlr import MultiVariableLinearRegression, find_best_mvlr, ValidationParameters
-from .models import IndependentVariable, MultiVariableRegressionResult
+from .main import find_best_mvlr
+from .models import (
+    IndependentVariableInput,
+    MultiVariableRegressionInput,
+    MultiVariableRegressionResult,
+    ValidationParameters,
+    IndependentVariableResult,
+)
 
 __all__ = [
-    "MultiVariableLinearRegression",
-    "MultiVariableRegressionResult",
-    "IndependentVariable",
     "find_best_mvlr",
+    "IndependentVariableInput",
+    "MultiVariableRegressionInput",
+    "MultiVariableRegressionResult",
     "ValidationParameters",
+    "IndependentVariableResult",
 ]

openenergyid/mvlr/main.py

@@ -0,0 +1,28 @@
+"""Main module for the MultiVariableLinearRegression class."""
+
+from .models import MultiVariableRegressionInput, MultiVariableRegressionResult
+from .helpers import resample_input_data
+from .mvlr import MultiVariableLinearRegression
+
+
+def find_best_mvlr(
+    data: MultiVariableRegressionInput,
+) -> MultiVariableRegressionResult:
+    """Cycle through multiple granularities and return the best model."""
+    for granularity in data.granularities:
+        frame = data.data_frame()
+        frame = resample_input_data(data=frame, granularity=granularity)
+        mvlr = MultiVariableLinearRegression(
+            data=frame,
+            y=data.dependent_variable,
+            granularity=granularity,
+            allow_negative_predictions=data.allow_negative_predictions,
+        )
+        mvlr.do_analysis()
+        if mvlr.validate(
+            min_rsquared=data.validation_parameters.rsquared,
+            max_f_pvalue=data.validation_parameters.f_pvalue,
+            max_pvalues=data.validation_parameters.pvalues,
+        ):
+            return MultiVariableRegressionResult.from_mvlr(mvlr)
+    raise ValueError("No valid model found.")

openenergyid/mvlr/models.py

@@ -1,5 +1,6 @@
 """Models for multivariable linear regression."""
-from typing import Optional
+from typing import Any, List, Optional
+import pandas as pd
 
 from pydantic import BaseModel, Field, ConfigDict
 import statsmodels.formula.api as fm
@@ -10,6 +11,118 @@
 from .mvlr import MultiVariableLinearRegression
 
 
+COLUMN_TEMPERATUREEQUIVALENT = "temperatureEquivalent"
+
+
+######################
+# MVLR Input Models #
+######################
+
+
+class ValidationParameters(BaseModel):
+    """Parameters for validation of a multivariable linear regression model."""
+
+    rsquared: float = Field(
+        0.75, ge=0, le=1, description="Minimum acceptable value for the adjusted R-squared"
+    )
+    f_pvalue: float = Field(
+        0.05, ge=0, le=1, description="Maximum acceptable value for the F-statistic"
+    )
+    pvalues: float = Field(
+        0.05, ge=0, le=1, description="Maximum acceptable value for the p-values of the t-statistic"
+    )
+
+
+class IndependentVariableInput(BaseModel):
+    """
+    Independent variable.
+
+    Has to corresponds to a column in the data frame.
+    """
+
+    name: str = Field(
+        description="Name of the independent variable. "
+        "If the name is `temperatureEquivalent`, "
+        "it will be unpacked into columns according to the variants."
+    )
+    variants: Optional[list[str]] = Field(
+        default=None,
+        description="Variants of the `temperatureEquivalent` independent variable. "
+        "Eg. `HDD_16.5` will be Heating Degree Days with a base temperature of 16.5°C, "
+        "`CDD_0` will be Cooling Degree Days with a base temperature of 0°C.",
+    )
+
+
+class MultiVariableRegressionInput(BaseModel):
+    """Multi-variable regression input."""
+
+    timezone: str = Field(alias="timeZone")
+    independent_variables: List[IndependentVariableInput] = Field(
+        alias="independentVariables", min_length=1
+    )
+    dependent_variable: str = Field(alias="dependentVariable")
+    frame: TimeSeries
+    granularities: list[Granularity]
+    allow_negative_predictions: bool = Field(alias="allowNegativePredictions", default=False)
+    validation_parameters: ValidationParameters = Field(
+        alias="validationParameters", default=ValidationParameters()
+    )
+
+    def model_post_init(self, __context: Any) -> None:
+        """Post init hook."""
+        # Check if all independent variables are present in the data frame
+        for iv in self.independent_variables:  # pylint: disable=not-an-iterable
+            if iv.name not in self.frame.columns:
+                raise ValueError(f"Independent variable {iv.name} not found in the data frame.")
+
+        return super().model_post_init(__context)
+
+    def _data_frame(self) -> pd.DataFrame:
+        """Convert the data to a pandas DataFrame."""
+        return self.frame.to_pandas(timezone=self.timezone)
+
+    def data_frame(self) -> pd.DataFrame:
+        """
+        Return the data frame ready for analysis.
+
+        Unpacks degree days and removes unnecessary columns.
+
+        If an independent variable named `temperatureEquivalent` is present,
+        it will be unpacked into columns according to the variants.
+        Eg. Variant "HDD_16.5" will be Heating Degree Days
+        with a base temperature of 16.5°C,
+        "CDD_0" will be Cooling Degree Days with a base temperature of 0°C.
+        """
+        frame = self._data_frame()
+        columns_to_retain = [self.dependent_variable]
+        for iv in self.independent_variables:  # pylint: disable=not-an-iterable
+            if iv.name == COLUMN_TEMPERATUREEQUIVALENT and iv.variants is not None:
+                for variant in iv.variants:
+                    prefix, base_temperature = variant.split("_")
+                    if prefix == "CDD":
+                        frame[variant] = frame[COLUMN_TEMPERATUREEQUIVALENT] - float(
+                            base_temperature
+                        )
+                    else:
+                        frame[variant] = (
+                            float(base_temperature) - frame[COLUMN_TEMPERATUREEQUIVALENT]
+                        )
+                    frame[variant] = frame[variant].clip(lower=0)
+                    columns_to_retain.append(variant)
+                frame.drop(columns=[COLUMN_TEMPERATUREEQUIVALENT], inplace=True)
+            else:
+                columns_to_retain.append(iv.name)
+
+        frame = frame[columns_to_retain].copy()
+
+        return frame
+
+
+######################
+# MVLR Result Models #
+######################
+
+
 class ConfidenceInterval(BaseModel):
     """Confidence interval for a coefficient."""
 
@@ -18,7 +131,7 @@ class ConfidenceInterval(BaseModel):
     upper: float
 
 
-class IndependentVariable(BaseModel):
+class IndependentVariableResult(BaseModel):
     """Independent variable for a multivariable linear regression model."""
 
     name: str
@@ -33,7 +146,7 @@ class IndependentVariable(BaseModel):
     model_config = ConfigDict(populate_by_name=True)
 
     @classmethod
-    def from_fit(cls, fit: fm.ols, name: str) -> "IndependentVariable":
+    def from_fit(cls, fit: fm.ols, name: str) -> "IndependentVariableResult":
         """Create an IndependentVariable from a fit."""
         return cls(
             name=name,
@@ -53,12 +166,12 @@ class MultiVariableRegressionResult(BaseModel):
     """Result of a multivariable regression model."""
 
     dependent_variable: str = Field(alias="dependentVariable")
-    independent_variables: list[IndependentVariable] = Field(alias="independentVariables")
+    independent_variables: list[IndependentVariableResult] = Field(alias="independentVariables")
     r2: float = Field(ge=0, le=1, alias="rSquared")
     r2_adj: float = Field(ge=0, le=1, alias="rSquaredAdjusted")
     f_stat: float = Field(ge=0, alias="fStat")
     prob_f_stat: float = Field(ge=0, le=1, alias="probFStat")
-    intercept: IndependentVariable
+    intercept: IndependentVariableResult
     granularity: Granularity
     frame: TimeSeries
 
@@ -73,7 +186,7 @@ def from_mvlr(cls, mvlr: MultiVariableLinearRegression) -> "MultiVariableRegress
         param_keys.remove("Intercept")
         independent_variables = []
         for k in param_keys:
-            independent_variables.append(IndependentVariable.from_fit(mvlr.fit, k))
+            independent_variables.append(IndependentVariableResult.from_fit(mvlr.fit, k))
 
         # Create resulting TimeSeries
         cols_to_keep = list(param_keys)
@@ -88,7 +201,7 @@ def from_mvlr(cls, mvlr: MultiVariableLinearRegression) -> "MultiVariableRegress
             r2_adj=mvlr.fit.rsquared_adj,
             f_stat=mvlr.fit.fvalue,
             prob_f_stat=mvlr.fit.f_pvalue,
-            intercept=IndependentVariable.from_fit(mvlr.fit, "Intercept"),
+            intercept=IndependentVariableResult.from_fit(mvlr.fit, "Intercept"),
             granularity=mvlr.granularity,
             frame=TimeSeries.from_pandas(frame),
         )

openenergyid/mvlr/mvlr.py

@@ -3,29 +3,12 @@
 
 import numpy as np
 import pandas as pd
-from pydantic import BaseModel, Field
 import statsmodels.formula.api as fm
 from patsy import LookupFactor, ModelDesc, Term  # pylint: disable=no-name-in-module
 from statsmodels.sandbox.regression.predstd import wls_prediction_std
 
 from openenergyid.enums import Granularity
 
-from .helpers import resample_input_data
-
-
-class ValidationParameters(BaseModel):
-    """Parameters for validation of a multivariable linear regression model."""
-
-    rsquared: float = Field(
-        0.75, ge=0, le=1, description="Minimum acceptable value for the adjusted R-squared"
-    )
-    f_pvalue: float = Field(
-        0.05, ge=0, le=1, description="Maximum acceptable value for the F-statistic"
-    )
-    pvalues: float = Field(
-        0.05, ge=0, le=1, description="Maximum acceptable value for the p-values of the t-statistic"
-    )
-
 
 class MultiVariableLinearRegression:
     """Multi-variable linear regression.
@@ -56,7 +39,6 @@ def __init__(
         confint: float = 0.95,
         cross_validation: bool = False,
         allow_negative_predictions: bool = False,
-        validation_params: ValidationParameters = None,
         granularity: Granularity = None,
     ):
         """Parameters
@@ -80,8 +62,6 @@ def __init__(
             If True, allow predictions to be negative.
             For gas consumption or PV production, this is not physical
             so allow_negative_predictions should be False
-        validation_params : ValidationParameters, default=None
-            Parameters to validate the model.
         """
         self.data = data.copy()
         if y not in self.data.columns:
@@ -95,7 +75,6 @@ def __init__(
         self.confint = confint
         self.cross_validation = cross_validation
         self.allow_negative_predictions = allow_negative_predictions
-        self.validation_params = validation_params or ValidationParameters()
         self.granularity = granularity
         self._fit = None
         self._list_of_fits = []
@@ -299,7 +278,7 @@ def remove_from_model_desc(x: str, model_desc: ModelDesc) -> ModelDesc:
         pars_to_prune = fit.pvalues.where(fit.pvalues > p_max).dropna().index.tolist()
         try:
             pars_to_prune.remove("Intercept")
-        except KeyError:
+        except ValueError:
             pass
         while pars_to_prune:
             corrected_model_desc = remove_from_model_desc(
@@ -310,7 +289,7 @@ def remove_from_model_desc(x: str, model_desc: ModelDesc) -> ModelDesc:
             pars_to_prune = fit.pvalues.where(fit.pvalues > p_max).dropna().index.tolist()
             try:
                 pars_to_prune.remove("Intercept")
-            except KeyError:
+            except ValueError:
                 pass
         return fit
 
@@ -400,40 +379,25 @@ def add_prediction(self):
         """
         self.data = self._predict(fit=self.fit, data=self.data)
 
-    @property
-    def is_valid(self) -> bool:
+    def validate(
+        self, min_rsquared: float = 0.75, max_f_pvalue: float = 0.05, max_pvalues: float = 0.05
+    ) -> bool:
         """Checks if the model is valid.
 
         Returns
         -------
             bool: True if the model is valid, False otherwise.
         """
-        if self.fit.rsquared_adj < self.validation_params.rsquared:
+        if self.fit.rsquared_adj < min_rsquared:
             return False
 
-        if self.fit.f_pvalue > self.validation_params.f_pvalue:
+        if self.fit.f_pvalue > max_f_pvalue:
             return False
 
         param_keys = self.fit.pvalues.keys().tolist()
         param_keys.remove("Intercept")
         for k in param_keys:
-            if self.fit.pvalues[k] > self.validation_params.pvalues:
+            if self.fit.pvalues[k] > max_pvalues:
                 return False
 
         return True
-
-
-def find_best_mvlr(
-    data: pd.DataFrame,
-    y: str,
-    granularities: list[Granularity],
-    **kwargs,
-) -> MultiVariableLinearRegression:
-    """Cycle through multiple granularities and return the best model."""
-    for granularity in granularities:
-        data = resample_input_data(data=data, granularity=granularity)
-        mvlr = MultiVariableLinearRegression(data=data, y=y, granularity=granularity, **kwargs)
-        mvlr.do_analysis()
-        if mvlr.is_valid:
-            return mvlr
-    raise ValueError("No valid model found.")