Fix function errors in bin_chi2 function.

rosaceae/bins.py

@@ -189,8 +189,7 @@ def chi2(a, bad_rate):
         chi = (a[0] - b[0])**2 / b[0] + (a[1] - b[1])**2 / b[1]
         return chi
 
-
-def recursion(groups, counts, bins, numeric=False, verbose=False):
+def recursion(groups, counts, bins, bad_rate, numeric=False, verbose=False):
     max_chi = 0
     if not numeric:
         for _i, i in combinations(range(len(groups)), 2):
@@ -220,16 +219,17 @@ def recursion(groups, counts, bins, numeric=False, verbose=False):
             elif chi_after > max_chi:
                 max_com_idx = (i, i+1)
                 max_chi = chi_after
+
         merged = (groups[max_com_idx[0]][0], groups[max_com_idx[1]][1]) # create a new boundary
         if verbose:
-            print(groups[max_com_idx[0]], groups[max_com_idx[1]], '-->' ,merged)
+            print('*',groups[max_com_idx[0]], groups[max_com_idx[1]], '-->' ,merged)
         groups = groups[:max_com_idx[0]] + [merged] + groups[max_com_idx[1]+1:]
         merged_counts = counts[max_com_idx[0]] + counts[max_com_idx[1]]
         counts = counts[:max_com_idx[0]] + [merged_counts] + counts[max_com_idx[1]+1:]
     if len(groups) <= bins:
         return groups
     else:
-        return recursion(groups, counts, bins, numeric=numeric, verbose=verbose)
+        return recursion(groups, counts, bins, bad_rate, numeric=numeric, verbose=verbose)
 
 
 def bin_chi2(xarray, y, bins, min_sample=0.01, na_omit=True, verbose=False):
@@ -242,8 +242,9 @@ def bin_chi2(xarray, y, bins, min_sample=0.01, na_omit=True, verbose=False):
         y = y[~pd.isna(xarray)]        
     elif not na_omit:
         out['Miss'] = xarray.index[pd.isna(xarray)]
-    total_bad = xarray.sum()
+    total_bad = y.sum()
     bad_rate = total_bad / len(y)
+    print('bad_rate', total_bad, len(y))
     # numeric or categorious
     if xarray.dtype == 'object':
         if verbose:
@@ -276,18 +277,10 @@ def bin_chi2(xarray, y, bins, min_sample=0.01, na_omit=True, verbose=False):
         for _i, b in enumerate(borders[1:]):
             start = borders[i_start]
             end = b
-<<<<<<< HEAD
             if sum((xarray >= start) & (xarray < end)) < len(y) * min_sample:
                 continue
             else:
                 tmp = y[(xarray >= start) & (xarray < end)]
-=======
-            if sum((xarray >= start) & (y < end)) < len(y) * min_sample:
-                continue
-            else:
-                print(_i, start, b)
-                tmp = y[(xarray >= start) & (y < end)]
->>>>>>> 226c7901c6fe634e021925973c741631ae9705ac
                 groups.append((start, end))
                 counts.append(np.array((sum(tmp), len(tmp)-sum(tmp))))
             i_start = _i + 1
@@ -301,7 +294,16 @@ def bin_chi2(xarray, y, bins, min_sample=0.01, na_omit=True, verbose=False):
 
         if verbose:
             print('Init groups:', groups)
-        groups = recursion(groups, counts, bins, numeric=True, verbose=verbose)
+        groups = recursion(groups, counts, bins, bad_rate=bad_rate, numeric=True, verbose=verbose)
+
+        # reformat numeric groups
+        for _i, g in enumerate(groups):
+            if np.isinf(g[0]):
+                g[_i] = '(-inf:%s]' % g[1]
+            elif np.isinf(g[1]):
+                g[_i] = '(%s:inf)' % g[0]
+            else:
+                g[_i] = '(%s:%s]' % (g[0], g[1])
 
     return groups
 
@@ -336,10 +338,10 @@ def bin_custom(xarray, groups, na_omit=True, verbose=False):
     # handle missing data.
     if not na_omit:
         if verbose:
-                print('Keep NA data')
+            print('Keep NA data')
         tmp = np.where(pd.isna(xarray))[0]
         if verbose:
-                print('Missing data: %s' % len(tmp))
+            print('Missing data: %s' % len(tmp))
         if len(tmp) > 0:
             out['Miss'] = np.where(pd.isna(xarray))[0]