mlquantify 0.1.0__py3-none-any.whl → 0.1.2__py3-none-any.whl

mlquantify/methods/threshold_optimization.py

@@ -1,5 +1,6 @@
 from abc import abstractmethod
 import numpy as np
+import warnings
 from sklearn.base import BaseEstimator
 
 from ..base import AggregativeQuantifier
@@ -446,9 +447,8 @@ class MS(ThresholdOptimization):
     {0: 0.3991228070175439, 1: 0.6008771929824561}
     """
 
-    def __init__(self, learner: BaseEstimator=None, threshold: float = 0.5):
+    def __init__(self, learner: BaseEstimator=None):
         super().__init__(learner)
-        self.threshold = threshold
 
     def best_tprfpr(self, thresholds: np.ndarray, tprs: np.ndarray, fprs: np.ndarray) -> tuple:
         """
@@ -481,11 +481,42 @@ class MS(ThresholdOptimization):
         ValueError
             If `thresholds`, `tprs`, or `fprs` are empty or have mismatched lengths.
         """
-        # Compute median TPR and FPR
-        tpr = np.median(tprs)
-        fpr = np.median(fprs)
 
-        return (self.threshold, tpr, fpr)
+        return (thresholds, tprs, fprs)
+
+    def _predict_method(self, X) -> dict:
+        """
+        Predicts class prevalences using the adjusted threshold.
+
+        Parameters
+        ----------
+        X : pd.DataFrame or np.ndarray
+            The input features for prediction.
+
+        Returns
+        -------
+        np.ndarray
+            An array of predicted prevalences for the classes.
+        """
+        # Get predicted probabilities for the positive class
+        probabilities = self.predict_learner(X)[:, 1]
+        
+        prevs = []
+        
+        for thr, tpr, fpr in zip(self.threshold, self.tpr, self.fpr):
+            cc_output = len(probabilities[probabilities >= thr]) / len(probabilities)
+            
+            if tpr - fpr == 0:
+                prevalence = cc_output
+            else:
+                prev = np.clip((cc_output - fpr) / (tpr - fpr), 0, 1)
+                prevs.append(prev)
+        
+        prevalence = np.median(prevs)
+        
+        prevalences = [1 - prevalence, prevalence]
+
+        return np.asarray(prevalences)
 
     
 
@@ -570,33 +601,63 @@ class MS2(ThresholdOptimization):
             - The median threshold value for cases meeting the condition (float).
             - The median true positive rate for cases meeting the condition (float).
             - The median false positive rate for cases meeting the condition (float).
-
+        
         Raises
         ------
         ValueError
-            If no cases satisfy the condition `|TPR - FPR| > 0.25` or if the 
-            input arrays are empty or have mismatched lengths.
+            If no cases satisfy the condition `|TPR - FPR| > 0.25`.
+        Warning
+            If all TPR or FPR values are zero.
         """
+        # Check if all TPR or FPR values are zero
+        if np.all(tprs == 0) or np.all(fprs == 0):
+            warnings.warn("All TPR or FPR values are zero.")
+        
         # Identify indices where the condition is satisfied
         indices = np.where(np.abs(tprs - fprs) > 0.25)[0]
         if len(indices) == 0:
-            raise ValueError("No cases meet the condition |TPR - FPR| > 0.25.")
+            warnings.warn("No cases satisfy the condition |TPR - FPR| > 0.25.")
+            indices = np.where(np.abs(tprs - fprs) >= 0)[0]
+            
+        thresholds_ = thresholds[indices]
+        tprs_ = tprs[indices]
+        fprs_ = fprs[indices]
 
-        # Compute medians for the selected cases
-        threshold = np.median(thresholds[indices])
-        tpr = np.median(tprs[indices])
-        fpr = np.median(fprs[indices])
-
-        return (threshold, tpr, fpr)
-
-    
-    
-    
-    
+        return (thresholds_, tprs_, fprs_)
 
+    def _predict_method(self, X) -> dict:
+        """
+        Predicts class prevalences using the adjusted threshold.
 
+        Parameters
+        ----------
+        X : pd.DataFrame or np.ndarray
+            The input features for prediction.
 
+        Returns
+        -------
+        np.ndarray
+            An array of predicted prevalences for the classes.
+        """
+        # Get predicted probabilities for the positive class
+        probabilities = self.predict_learner(X)[:, 1]
+        
+        prevs = []
+        
+        for thr, tpr, fpr in zip(self.threshold, self.tpr, self.fpr):
+            cc_output = len(probabilities[probabilities >= thr]) / len(probabilities)
+            
+            if tpr - fpr == 0:
+                prevalence = cc_output
+            else:
+                prev = np.clip((cc_output - fpr) / (tpr - fpr), 0, 1)
+                prevs.append(prev)
+        
+        prevalence = np.median(prevs)
+        
+        prevalences = [1 - prevalence, prevalence]
 
+        return np.asarray(prevalences)
 
 class PACC(ThresholdOptimization):
     """

mlquantify/model_selection.py

@@ -131,7 +131,7 @@ class GridSearchQ(Quantifier):
                  model: Quantifier,
                  param_grid: dict,
                  protocol: str = 'app',
-                 n_prevs: int = None,
+                 n_prevs: int = 100,
                  n_repetitions: int = 1,
                  scoring: Union[List[str], str] = "ae",
                  refit: bool = True,

{mlquantify-0.1.0.dist-info → mlquantify-0.1.2.dist-info}/METADATA

@@ -1,6 +1,6 @@
-Metadata-Version: 2.2
+Metadata-Version: 2.4
 Name: mlquantify
-Version: 0.1.0
+Version: 0.1.2
 Summary: Quantification Library
 Home-page: https://github.com/luizfernandolj/QuantifyML/tree/master
 Maintainer: Luiz Fernando Luth Junior
@@ -58,7 +58,7 @@ pip install mlquantify
 If you only want to update, run the code below:
 
 ```bash
-pip install --update mlquantify
+pip install --upgrade mlquantify
 ```
 
 ___

{mlquantify-0.1.0.dist-info → mlquantify-0.1.2.dist-info}/RECORD

@@ -1,6 +1,6 @@
 mlquantify/__init__.py,sha256=Q9jCEkG0EJoHXukrxh194mhO_Yfu-BZPRfjpQ4T1XlQ,978
 mlquantify/base.py,sha256=hJ9FYYNGeO5-WJlpJpsUiu_LQL1fimvZPPNsKptxN7w,19196
-mlquantify/model_selection.py,sha256=deMdTKarMu4DtfSy6ucTIxQKj1yrC-x3nhBT77kMOtI,12679
+mlquantify/model_selection.py,sha256=rPR4fwwxuihzx5Axq4NhMOeuMBzpoC9pKp5taYNt_LY,12678
 mlquantify/plots.py,sha256=9XOhx4QXkN9RkkiErLuL90FWIBUV2YTEJNT4Jwfy0ac,12380
 mlquantify/classification/__init__.py,sha256=3FGf-F4SOM3gByUPsWdnBzjyC_31B3MtzuolEuocPls,22
 mlquantify/classification/methods.py,sha256=yDSbpoqM3hfF0a9ATzKqfG9S-44x-0Rq0lkAVJKTIEs,5006
@@ -12,11 +12,11 @@ mlquantify/methods/aggregative.py,sha256=rL_xlX2nYECrxFSjBJNlxj6h3b-iIs7l_XgxIRS
 mlquantify/methods/meta.py,sha256=sZWQHUGkm6iiqujmIpHDL_8tDdKQ161bzD5mcpXLWEY,19066
 mlquantify/methods/mixture_models.py,sha256=si2Pzaka5Kbva4QKBzLolvb_8V0ZEjp68UBAiOwl49s,35166
 mlquantify/methods/non_aggregative.py,sha256=xaBu21TUtiYkOEUKO16NaNMwdNa6-SNjfBsc5PpIMyI,4815
-mlquantify/methods/threshold_optimization.py,sha256=P88VXG-czZiaHSHTGnzFmZVzm3SoJHnrmi60Zvv7IJU,33726
+mlquantify/methods/threshold_optimization.py,sha256=-iOcP5YcXZd0XZHGvbmcoE72hXR6D9YCoTnr1l80-9k,35796
 mlquantify/utils/__init__.py,sha256=logWrL6B6mukP8tvYm_UPEdO9eNA-J-ySILr7-syDoc,44
 mlquantify/utils/general.py,sha256=Li5ix_dy19dUhYNgiUsNHdqqnSVYvznUBUuyr-zYSPI,7554
 mlquantify/utils/method.py,sha256=RL4vBJGl5_6DZ59Bs62hdNXI_hnoDIWilMMyMPiOjBg,12631
-mlquantify-0.1.0.dist-info/METADATA,sha256=6ud6gvzxxaQr7oZLD3fu3piid1ZHjJuHAQZzZeUw7Rs,4939
-mlquantify-0.1.0.dist-info/WHEEL,sha256=In9FTNxeP60KnTkGw7wk6mJPYd_dQSjEZmXdBdMCI-8,91
-mlquantify-0.1.0.dist-info/top_level.txt,sha256=tGEkYkbbFElwULvqENjam3u1uXtyC1J9dRmibsq8_n0,11
-mlquantify-0.1.0.dist-info/RECORD,,
+mlquantify-0.1.2.dist-info/METADATA,sha256=2j3pqrm5djMAPm7bKTIjBjtg71OzAbFpwC-_ofOoSlc,4940
+mlquantify-0.1.2.dist-info/WHEEL,sha256=CmyFI0kx5cdEMTLiONQRbGQwjIoR1aIYB7eCAQ4KPJ0,91
+mlquantify-0.1.2.dist-info/top_level.txt,sha256=tGEkYkbbFElwULvqENjam3u1uXtyC1J9dRmibsq8_n0,11
+mlquantify-0.1.2.dist-info/RECORD,,

{mlquantify-0.1.0.dist-info → mlquantify-0.1.2.dist-info}/WHEEL

@@ -1,5 +1,5 @@
 Wheel-Version: 1.0
-Generator: setuptools (75.8.0)
+Generator: setuptools (78.1.0)
 Root-Is-Purelib: true
 Tag: py3-none-any