bug fix for rate tables

ssda903/multinomial.py

@@ -4,7 +4,7 @@
 
 import numpy as np
 import pandas as pd
-from demand_model.base import BaseModelPredictor, combine_rates
+from demand_model.base import BaseModelPredictor
 from demand_model.multinomial.utils import (
  build_transition_rates_matrix,
  fill_missing_states,
@@ -31,6 +31,23 @@ class NextPrediction:
  variance: np.ndarray
 
 
+def combine_rates(rate1: pd.Series, rate2: pd.Series) -> pd.Series:
+ """'
+ This has been updated to a multiplication method.
+ Fill value=1 allows any values missing from the adjustment series to remain unchanged in the transition rates
+ Any rates not present in rate1 will not be included in output
+ """
+ rate1, rate2 = rate1.align(rate2, fill_value=1)
+
+ # Create a mask to identify where rate1 is missing but rate2 is not
+ mask = (rate1 != 1) | (rate2 == 1)
+
+ # Apply the mask to exclude undesired cases
+ rates = (rate1 * rate2)[mask]
+ rates.index.names = ["from", "to"]
+ return rates
+
+
 class MultinomialPredictor(BaseModelPredictor):
  def __init__(
  self,