From f38403f0d661b91ee2a8c0192f8bd01ed4e0c05e Mon Sep 17 00:00:00 2001
From: zulfikar yuniar <zulfikaryuniar06@gmail.com>
Date: Wed, 4 Jun 2025 23:11:40 +0700
Subject: [PATCH] revisi algoritma

---
 .../decision_tree_model.cpython-310.pyc       | Bin 3965 -> 4315 bytes
 Fastapi/decision_tree_model.py                |  27 +++++++++++++++---
 2 files changed, 23 insertions(+), 4 deletions(-)

diff --git a/Fastapi/__pycache__/decision_tree_model.cpython-310.pyc b/Fastapi/__pycache__/decision_tree_model.cpython-310.pyc
index 2a4b9425182a4de050a967bb25fb399cebb7d751..a89df2357de0a9090ad5a9a8c3f360b34ffd43c2 100644
GIT binary patch
delta 1632
zcmZ8h&2Jk;6rY*>@OZtuc4AjaNYap|bz%yrQ36Uy0%;o>TB$8*KuA`KyJX$kj$^yC
zn+Bv=MaZdCA*w2w9(zc+^$I8YCvfTAy}%I=;(|CJmGa)YDOGnhzxO$Be)D#Ao*sTO
z>NXw6BJljSu~7M`ch?=F_jZqeJTT<jzOzG%j_>;E9a41tAwT2y?9ko4G3PaZ5L`K*
zu8_j;?xV+tr_a2%TK9x^z1djZ+S*$E<kO{7E2mb^dO_&5sy}>3Bc^`0occ5@BuPjH
z$Uc2QA{sJD+SI4(tT_;BlGb#vSfy(-NxLyvMZ5SbgSjFHP%uOabsGV>ceJ1!cFdbz
zeX|@Cm~b(WR=ZYqJ5vox%|^M@DvQ!u-3x*~hy^-;O(&=r0c8yfA3Ru}!#D#Y*rULf
z`ec)cL`d4mmm=SYb$LgcWXL^0?zj)OPQ<9Mg}T(*RBFDyeP3H4RTAki3`uInBDwNe
zWJ)tMVcMKDCA~?4xiGQMC6VR;sU^ci&3eFs`5O~m&wdITl>j}C%fvbzAQ5IC*gDur
zAjW~11u_XT1%$v3w_{#|B(lK)fi6$59cfD|N2DWDInqZWS2F4LlPE3IQkUi&kqK#_
z1R8qq&C7ejErm8@={3hU?yx8WnHeCFkW^+W^fZZjVtU}a`kZg?fdg#hn+;vr+or+y
zAhT^M9=*DNZ^{f%(89Frl|5|+GK<nr+&AGk@QVVz4G}}?wsCT56x2@A3qw);Vlymv
zl3_)ZgG#gRi!lu8_42i_V~FbYN?6E<VeqLxjb$VL{A$7M>(6y`-wVBBdNU~dr5BX~
z^|_g+d3DQd<+6?PR_WF7U*-|D!1e7RC>dM>;MAlX-W+4hq>P$RbyKGnc))tr%h<n6
zvuOri&N$65&_Hr#QC&Nj;hG#~#IYAXmS9kg#kI|U#0gmHrE{ykPOOb!7sTrbM-Yx7
z<PpXZCX~htld+vZI|e-IVhZ3Yi~xTdJ6?fPo4V9xVoF`%H@cONLbO}7TMx{F-tFZd
ze4I|JL&=fh8SvI%#I<nGaUH0|shCmA$-8t?{gTW>Hn)l<?gHEuYs|!pLCnH0J5B@j
zqcuvG)D!ES^engwR4k!jVo{Bx-iFHv>f;}beHQ1`YHIcvPKN}{i}zE|#3JZLtJ-L7
zh9$qc7K)GXLQs!W{ps5fdS1b$;=KAh^^o3C`}Pc<?ADoAKiemrZ!rxoha44`RNh$z
z@?Sc|ci#Y)IF9fp!dn2awgIA}tu^ZpsW^-N|2upE%ss{33n%6=@IJx^2&WP7su2YM
z+#%{8BE++o)h#!_eHZKi?>QHa1soJ=J`<NQ2b)khR5V*&!}o%?TG421ZnVA<$g!hu
zc<?7^m>Mq}-<or1_8?>t1`+TYjL%fuJb_=a+skvH+2TCfmk=(i$#j0=D5xCTkB3EP
XIv!=tHLh6(U$iWq<UL&Hx~=^O<se5u

delta 1282
zcmZ8h&2Jk;6rY*>@OZr+PGTt^Z5>P+mjyK>2$`lSh@=Urgz}|;kgQb3WZlMzn>e#$
zqM%v%&>Rp_1<`Qqp}kRYKvlVO;l!olh~dJKzn~JRg7?;;NO&~A_kJ^P-~Q&!TmM>m
zUi1$<&mr(z{^gyWXThm|o_=xq+WN2?c_*amMPrd41t+xXNBL;rgq(goVbM(zo;&^V
z>FDd%uWq&KqJH~ef9vSzXzPOym*+R<w-%P3Jf;cj4no!p%<C}YRZ55?irl3Ml{B?X
z(TMJ`cu+Fr5!)kTEHb2_*j=hvWZpM6Ns}Z7OjD6vbCKNKOe|$d3!a*h_JfQf${HoR
zT-v+Yhb(^Mu+(39;5;PV#0HWED0`2Nk_5y5<tA_rf-)c|3zma*KnU=@oO2|}fd}g^
zxT|t1TOi6+PJs-Q#8XsxBP8*auMA~fBg$4LT99GH-R}R-B6ARC?7ES4k0k+w1|Xs@
z61+2!`BY?|qsW2CRsket(o=a5<Ea6tRp!18x4>5gl$HF!0D5MtQNI-$JyRaH8$GA3
zzqKc6QU7d@4ZU=>e@pf((Y(DQLth)2jbg9xNvGM0YEm>BwOESIw(N*TIM_3zx~%KZ
zGi5rVzst1so4j=MG9(#>8Lt9x%4vZyW>ZG(=ca8^hn@%DrFj^fahhk~f#uAh=8H(*
z?(Z;zK{Pyz<4<u}|HKc+i;!4cKp01O8DRq9BEqDuX2Y43V^@h3IC|3nD==dpzy)PE
zT<TMwiD~_P_AGM=W7DMld_Y69XT%Mu7oAIVMz@@Cs{*SZ>93q&TG8J)U(zXkC0B;M
zTCQqi1w=*vkeje)(4E!4vmyOgu0TK5dACHDbkV$^*W8g{7IRavgocR){i!<@90G|=
z08}TI^%Hk)64O8c#-*nS4Y2@z)oJdxJF*rvx21S5eM{LJ2|BQKmcgguhF<X=(a-cV
z?+Tyl=b6)Q`Lo_JhG9z-s94d7zX6AM>Q^h5U`f1!5F)$^0PnWI^o;F;)*~twas9u#
zz7MpkYr)F&9By1gcpKqegn5KA0G<$i{LcGNuj$`{@|pr3<9~|}%>hk7?Z3r3#$XY`
zLe*~9_oI58W~*BJojdJgff{?}ox0o+28O0D&Y^czZ{&wArdNRXE%2iHg<D4N1_CyL
iSl9RR<ucwghf-rZn9RhM<lKN!aU7oI1Ki}MYy1PBCICSI

diff --git a/Fastapi/decision_tree_model.py b/Fastapi/decision_tree_model.py
index 65f1cf8..dc3adbb 100644
--- a/Fastapi/decision_tree_model.py
+++ b/Fastapi/decision_tree_model.py
@@ -20,7 +20,14 @@ def count_classes(subset):
 def get_majority_class(counts):
     return max(counts.items(), key=lambda x: x[1])[0]
 
-def build_tree_structure(data):
+def is_one_per_class(subset):
+    counts = subset[target_column].value_counts()
+    return len(subset) == 2 and set(counts.values) == {1} and len(counts) == 2
+
+
+def build_tree_structure(data, used_attributes=None):
+    if used_attributes is None:
+        used_attributes = set()
     entropy = calculate_entropy(data)
     counts = count_classes(data)
 
@@ -34,7 +41,9 @@ def build_tree_structure(data):
     gains = {}
     thresholds = {}
     for column in data.columns:
-        if column == target_column or not np.issubdtype(data[column].dtype, np.number):
+        if (column == target_column or
+            column in used_attributes or
+            not np.issubdtype(data[column].dtype, np.number)):
             continue
         threshold = data[column].mean()
         le_subset = data[data[column] <= threshold]
@@ -51,6 +60,7 @@ def build_tree_structure(data):
 
         gains[column] = gain
         thresholds[column] = threshold
+       
 
     if not gains:
         return {
@@ -61,17 +71,26 @@ def build_tree_structure(data):
 
     best_attribute = max(gains, key=gains.get)
     threshold = thresholds[best_attribute]
+    new_used_attributes = used_attributes | {best_attribute}
 
     le_data = data[data[best_attribute] <= threshold]
     gt_data = data[data[best_attribute] > threshold]
 
+    #pengecekan jika node diteruskan akan menghasilkan leaf node kelas 1=1 dan 0=1
+    if is_one_per_class(le_data) or is_one_per_class(gt_data):
+        return {
+            "type": "leaf",
+            "prediction": get_majority_class(counts),
+            "class_counts": counts
+        }
+
     return {
         "type": "node",
         "attribute": best_attribute,
         "threshold": threshold,
         "class_counts": counts,
-        "left": build_tree_structure(le_data),
-        "right": build_tree_structure(gt_data)
+        "left": build_tree_structure(le_data, new_used_attributes),
+        "right": build_tree_structure(gt_data, new_used_attributes)
     }
 
 def predict(tree, input_data):