From 58920c82c27bbe7797dd3b363eb62f9061b854ef Mon Sep 17 00:00:00 2001
From: Valentin Belyaev <chuvalik.work@gmail.com>
Date: Tue, 6 Feb 2024 16:33:06 +0300
Subject: [PATCH] Add R2 metric + test

---
 .gitignore                                    |   2 +-
 setup.py                                      |   2 +-
 .../__pycache__/__init__.cpython-310.pyc      | Bin 220 -> 220 bytes
 .../__pycache__/boosting.cpython-310.pyc      | Bin 5361 -> 5361 bytes
 .../__pycache__/xgboost.cpython-310.pyc       | Bin 9261 -> 9374 bytes
 src/ensemble/catboost.py                      |  56 ++++++++++++++++
 .../classification.cpython-310.pyc            | Bin 3648 -> 3648 bytes
 .../__pycache__/regression.cpython-310.pyc    | Bin 895 -> 1350 bytes
 src/metrics/regression.py                     |  23 +++++--
 ...test_r2_score.cpython-310-pytest-7.4.4.pyc | Bin 0 -> 2410 bytes
 tests/metrics/regression/test_r2_score.py     |  63 ++++++++++++++++++
 11 files changed, 140 insertions(+), 6 deletions(-)
 create mode 100644 src/ensemble/catboost.py
 create mode 100644 tests/metrics/regression/__pycache__/test_r2_score.cpython-310-pytest-7.4.4.pyc
 create mode 100644 tests/metrics/regression/test_r2_score.py

diff --git a/.gitignore b/.gitignore
index 3a0b019..1eddeb1 100644
--- a/.gitignore
+++ b/.gitignore
@@ -3,4 +3,4 @@
 src/__pycache__
 build
 dist
-tulia.egg-info
\ No newline at end of file
+tulia.egg-info
diff --git a/setup.py b/setup.py
index 754278e..f9064da 100644
--- a/setup.py
+++ b/setup.py
@@ -1,6 +1,6 @@
 from setuptools import setup, find_packages
 
-VERSION = '0.2.1'
+VERSION = '0.3.0'
 DESCRIPTION = 'numpy based machine learning package with sklearn-like API'
 
 with open("README.md", "r") as fn:
diff --git a/src/ensemble/__pycache__/__init__.cpython-310.pyc b/src/ensemble/__pycache__/__init__.cpython-310.pyc
index 2c38bf5c08389fccfe149be4cb9d2fd75d661c71..be836631b94345015a6fa034aeeece191987aba4 100644
GIT binary patch
delta 18
Ycmcb^c!!ZIpO=@50SJ^GCvsf^04a0?DF6Tf

delta 18
Ycmcb^c!!ZIpO=@50SH3nCURW@04d@GH~;_u

diff --git a/src/ensemble/__pycache__/boosting.cpython-310.pyc b/src/ensemble/__pycache__/boosting.cpython-310.pyc
index d8f4cb4142de3da9aaf234914726db1f2df2a234..cdb0c24847a3e8397c99d2939bfe65ca03d3aa2b 100644
GIT binary patch
delta 19
ZcmeyU`B9TApO=@50SJ^GH*!4}0RT3i1nd9+

delta 19
ZcmeyU`B9TApO=@50SIhuH*!4}0RT5A1p)v7

diff --git a/src/ensemble/__pycache__/xgboost.cpython-310.pyc b/src/ensemble/__pycache__/xgboost.cpython-310.pyc
index 47242cb845ade29b0efdb2eac1462c6369bdd81a..890dc234061b401c6fc03d9fa3799673b6069069 100644
GIT binary patch
delta 888
zcmZvZKWNlo7{>GFlH4VCxi-E3ZLiJURjRZW1*w9A)Wr&(+UlUE^gR21=|44)oFG@U
z{^?Y3>9=)oadL2wh&TvNZgvnsp}2O|!C8F27<yVv_#X28z9)IZ^SqpXcGh%F(~#(@
zO}$|sUYW1eH4P+*New`$AtN+mx~#`^g>Dr`P-6<lX(gD7IxT~#j)6%`t*Z?UO%`8H
zU|Oi7U=Q}LD-AuORz&+SL)7@&i`no!iz!THYpN4-5}3n0*^>X*3N+B7fe9>PiB414
z&rBj|8a9A~M0)lOE8`H+-iVT$oWp1nv;O=hz1=5<(_qDkl_W>GfEBEgA%nF`(xC%8
z)|E4-XW4?se1FNkbJ=68Gj7;)b)nrRYCEoHlh<zDT5!Fk_Tue{&=!2t?#eJ3%<7#O
z)PwERM}YgmR{Dty8^Ms(l;L^sAhQYZHn4K50H68O{5{wXv_c2q3x8Dn1-txVX{wa<
zvec886Ot#{i8q_*&NrL<Q^_}u5lKi2$QnP_Utim#?SD#Cb&c*grInx(qXIEWxu$4u
z(n?bf>99IcXjmGC2~lzx%u+CVL~~K0c`Oh$_Gps$MfyAy6$aWZi}D=YNp<sdz0PQu
z5h2(0+hK&>Ue<QI<*)criYEtN$k5`u<?~ugByf{g^ds;ySRPu1Bnom+$UOfsJemw0
zSH+(3syz*T!HS&+m<S$M%*f5eX^U@FM_`e^uNL4q|5`n)H3h!Gjaqrk+?Tg^O`H{l
r6om8$8544y*K6aMIePp~%bz5Ps@SL<g*pDNc7k+9E%4Nc4X1ts+K$r1

delta 835
zcmZ8eON$dh5T2e#GVeSm&rO`&3F@M&xOiCbAnL)*!MBKxtjTJ17Cji5AZQlJ!5|3Y
z!M^Lsg9kzB#gpJ!SP(@od-LSM{Re_tJq}SbO?^~-RXx>J-_!QXWvA~rri9<O!*A$?
z*UpAEl8GcyDI_3;vOx^lQ1GY_fJPNap{1z`I+{V%o(U#cV=dIcrl}iga0K>La3M2R
zLp@=$37dmFvc|u)P>AnksX|r!pVAxYPy`Rl%>T+t*f5I)I+US;SqrMvM#jNrHK-%w
zPBoi>2C|ui#cHf<!VcuJf1ZiAbD}vmR(eWR!n_Z&;!+;w&PnsFt=z@qN6*r=hz<tV
z`!~-=lyWzHTPBNqYz!6B=D+MsLhkZS_r6RX@YA`DOrG&qg(rk;@MGROAs^Vg(jD@i
zFO`Rcd}6%v)B1|rjzWZCjjehf`&gaJ7NVkHJpm;IX4eM#B09PxE9_zI#ON^wQ>v(j
zh9`yNr%{ze5ioI47REN3gM+!T7DAy|1&Bw+QS(s1N%e#+CJK2_Le`kDX1pMU1}IIu
zp~?1fLG>;B?}GXfe_nqflP){eJelg^bE8$(@f+-xQ6Ss=b90?o*95y;z*QFdyR5kC
zl8Avl@|TIpzxX9WcJl36C#eRJ?J_lJRo8^;Ou!Z1K7{@3NU+epBEpV<nt-~1nF%z|
yo42R7992ZVEFdFbkATbUZLl}Dg6?;;!9j#rgdf2HvceWxONeK*GP~FE$)Vqu)VRa|

diff --git a/src/ensemble/catboost.py b/src/ensemble/catboost.py
new file mode 100644
index 0000000..cb49ad1
--- /dev/null
+++ b/src/ensemble/catboost.py
@@ -0,0 +1,56 @@
+from typing import Union
+
+import numpy as np
+
+from src.base import Model
+
+
+# 1. Ordered Target encoding
+# 2. Boostrap data
+# 3. Symmetric tree
+# 4.
+
+
+class CatBoostClassifier(Model):
+    """
+    CatBoost for classification tasks.
+    """
+
+    def __init__(
+            self,
+            learning_rate: float = 3e-1,
+            n_steps: int = 100,
+            max_depth: int = 3,
+            cat_features: list = None
+    ):
+        pass
+
+
+    def fit(self, x: np.ndarray, y: np.ndarray):
+        pass
+    def predict(self, x: np.ndarray) -> np.ndarray:
+        pass
+
+    def _predict(self, x: np.ndarray) -> Union[np.ndarray, float, int]:
+        pass
+
+    def _encode_cat_features(self, x: np.ndarray, y: np.ndarray) -> np.ndarray:
+        cat_feature_idxs = []
+
+        encoded_features = []
+        for idx in cat_feature_idxs:
+            encoded_feature = []
+            option_count = {}
+            total_count = {}
+            for i, x_sample in enumerate(x[:, idx]):
+                ctr = (option_count.get(x_sample, 0) + 0.05) / (total_count.get(x_sample, 0) + 1)
+                encoded_feature.append(ctr)
+
+                if y[i] == 1:
+                    option_count[x_sample] = option_count.get(x_sample, 0) + 1
+                total_count[x_sample] = total_count.get(x_sample, 0) + 1
+
+            encoded_features.append(np.array(encoded_feature))
+
+        x[:, cat_feature_idxs] = encoded_features
+        return x
diff --git a/src/metrics/__pycache__/classification.cpython-310.pyc b/src/metrics/__pycache__/classification.cpython-310.pyc
index 974888af35980b83ddae0feee4cf6ff4d0ab2c99..ce7a319f9639045a867bada467ea711f165b9744 100644
GIT binary patch
delta 52
zcmX>gb3leWpO=@50SJ^Gw{7IM<YkLuE-5IAn(W0}%xE}yAFn%`AyBf2b8;c8@MJkY
FB>+&S3~2xW

delta 52
zcmX>gb3leWpO=@50SGKFY}v?d$;)<&xul@z)?_c<Vn*Z1`*_{ijDeCxoRbS#g(u7L
GDFFb5v<?se

diff --git a/src/metrics/__pycache__/regression.cpython-310.pyc b/src/metrics/__pycache__/regression.cpython-310.pyc
index 5f467cb5554972732d7c2800ca78fd3c32a156d3..55b66958c1cfb2732b20ae8d1df3544aa82359e2 100644
GIT binary patch
delta 491
zcmZus!AiqG5Z&2LW11!{*n_1OjTdXMdJs=J2^DW5=*eD6=&lE?ZJkX)N=h&OK*$eR
z{RA(bJ@^sgzvylnD3lJ&o1OP&hc|!bw_P5(u1!>k(=TyrB;^KllEbrID&U@SmZ2fo
zoym*bela>Gha2LTlF<Xqlizm5I~a7ubND#w3)%AgG@;E^?&AVXfVcFLp07&l#?w!V
zvmVu+@+hcks%)#2fr5y<LadR-WDYgLp{nYXF8FxKFiv9@(``~CQzJH%b2fRmJr(V+
z-y8SCNO;#r@@X8R;K!K|j6w|43Ef)fb+&SD{sl!e#zC-L<;mU<#X?1G9rX=9LrcGl
z8Y-G7puQ8Cf*%%1*<q-lh~3;0I$F-a)iZk*o^-R_4b@7$t!8b4Lk+;HYW$6~#hOq}
Quh_n$Rg2VF-Q6ku1Mv}YtN;K2

delta 86
zcmX@c^`DI|pO=@50SKm+uTGuJG?7n+(P5&t11n=NgC^(1u58BUiQk+jdoZeRu480l
mWNew-%H#s#PUd4SV$x)r{Dwt+vIMIV7aPzB4hAV^IbHzl9~8s@

diff --git a/src/metrics/regression.py b/src/metrics/regression.py
index 8f70c78..b91419e 100644
--- a/src/metrics/regression.py
+++ b/src/metrics/regression.py
@@ -4,8 +4,8 @@
 def mean_squared_error(y_true: np.ndarray, y_pred: np.ndarray) -> float:
     """
     Calculate mean-squared error.
-    :param y_true: Target labels.
-    :param y_pred: Target predictions.
+    :param y_true: Target labels (n_examples, ).
+    :param y_pred: Target predictions (n_examples, ).
     :return: Loss.
     """
     n_examples = len(y_true)
@@ -17,9 +17,24 @@ def mean_squared_error(y_true: np.ndarray, y_pred: np.ndarray) -> float:
 def mean_absolute_error(y_true: np.ndarray, y_pred: np.ndarray) -> float:
     """
     Calculate mean-absolute error.
-    :param y_true: Target labels.
-    :param y_pred: Target predictions.
+    :param y_true: Target labels (n_examples, ).
+    :param y_pred: Target predictions (n_examples, ).
     :return: Loss.
     """
     error = np.mean(np.abs(y_true - y_pred))
     return error
+
+
+def r2_score(y_true: np.ndarray, y_pred: np.ndarray) -> float:
+    """
+    Calculate R-squared.
+    :param y_true: Target labels (n_examples, ).
+    :param y_pred: Target predictions (n_examples, ).
+    :return: R-squared score.
+    """
+
+    tss = np.sum((y_true - np.mean(y_true))**2)
+    rss = np.sum((y_true - y_pred)**2)
+
+    r_squared = 1 - rss / tss
+    return r_squared
diff --git a/tests/metrics/regression/__pycache__/test_r2_score.cpython-310-pytest-7.4.4.pyc b/tests/metrics/regression/__pycache__/test_r2_score.cpython-310-pytest-7.4.4.pyc
new file mode 100644
index 0000000000000000000000000000000000000000..d9fb0d82f4432bfe91093313ca697f2901f910a4
GIT binary patch
literal 2410
zcmbtV&2Jk;6yII%de>{mDT0QUuSS7FjHs13X%i5wiz>tgMHL7Mt&CP1&&1hk?X}*l
zQ!LgOiUb$_01$_mn{OQ8j<_Mg4GC#aNUg9|>H*XX>H+cItev<bQ6pun`OSMX&ol31
zezW1+oULH{dHp9opI4N-q|7!AmFw8WHUd+a8Yqq`N3EUH6jf>Kj-h5NdB?;REwI`J
zr+{-grhle5HZxcr$09SCg=2{on2qBcE3y)fWj4pkIL<TmeWfy=;yKDzMNKTIxuO{Y
zZ_>=P6;L3LZ(tYyB1Fo*+Csm+Rfp=}QB7sqK#R0kiPbyWAQvgOas!>^1_t(gtjAiM
zi;XzX^e^>+8JnU<o<{n<F;w;xQH%2XW~2;N*g(l5siWdL%CSCFnR#IO4q3(;TD~h+
zQ9?_M_Q2vpX^}&VkE8`kXkkg&7^-s2Pev2>E_`DyC<7}h;EEkvL-mgODP~<)20C^F
zdwzhq#(ddZBO_HY4=GhD%@ODX2ErISw!iA*zjyC`@xwp&u8lND_o6UxbU=LA^@QLM
zEnZyi_E%Rb!v16`EnbzAwdouRlPNT-`l1<xg4=^-vR|u+*QE723AFuHn!Z9B9V4~P
zim2FwMC){uBLBF{n-OO&)DjJ9Bh$%s`zr)Bf;ED5f-3}@j!xc<iZF-4B!&mz^?|}o
z%!wmO>3X;CZuk5k@;f5Q2Vv6-L}KG?#6{$KkeqYHZrBU(Ox^{z6@=Sf;C8$=PYSN^
zb~sHZ<+o%${IK&jKnTf`?oJ48FLKf2!0UJsu8ay!zVAlR<1}`GGbp1kFpuCA$&W0l
ztFy}GS!H8ZdEJq|XXmW0judzfH+uNqjW-$}U~NQW&kJ~`akJmt^`Lzd!dsX+(P#&k
z+9BhCXi$Kt^9cN=XaH{k7Xtm13z*@IMAdFTIVW?0Gd|;;$j206VlbBiiLX2$R+ZJV
zs%y6Tn+7kSdMxuE9LqdS@=!&F{hxU{GEaYmdBy?r@-uuont7&V2{G@GS4T6?Vui;r
zk9cLpM={SnG4qOBKy0fRKvV<DA(#XfLQ3j{et*<~atlR*5&_-g!%G8aNPU9fSpuqm
z2G66EEKI8=cKsb9O;$7crwe5IBEqqif<;`d9QsPmWo}A`rR(%ww6J5dems@0PfvXX
zzK3`Bz}aVTYmm;qgwkW3tr#g~EPaZAE@HY&P(#4ij)<0?C!omFO9UGTlMwEoF5PFR
zAp{<@xZCNqw=oiOI}2b1r4s}IOK3M<lkIMOyd&G6_3_tVQA_LNUw@%?zmE53l{Tl~
zV2xm%fWC8bk5knnV|-h1m+_|GrWY@Zr%Kz)f?P)FB=KCNf4ofa3V|%h=On#~(E9Ge
z+4A>Gzn?Dk*-a;JZ~M`%8}_2W$J}qyDbiP}a^*~7O&@xo3zD37vsb>Fy}S7IRe|3I
zKf<T|U@9>ubFe`FF)$C^z9gbk6ws_r-cR{ik#C=*709|%$$|0&<iHF+B~v9ytvB&C
X?gjiBeToHfb4e*_R?f1sU&;OtK=?S2

literal 0
HcmV?d00001

diff --git a/tests/metrics/regression/test_r2_score.py b/tests/metrics/regression/test_r2_score.py
new file mode 100644
index 0000000..c6978d5
--- /dev/null
+++ b/tests/metrics/regression/test_r2_score.py
@@ -0,0 +1,63 @@
+import numpy as np
+
+from src.metrics import r2_score
+
+
+def test_r2_identical_arrays():
+    y_true = np.array([1, 2, 3, 4, 5])
+    y_pred = np.array([1, 2, 3, 4, 5])
+
+    expected_r2 = 1.0
+    r2 = r2_score(y_true, y_pred)
+
+    assert np.isclose(expected_r2, r2, atol=1e-5, rtol=1e-5)
+
+
+def test_r2_shifted_arrays():
+    y_true = np.array([1, 2, 3, 4, 5])
+    y_pred = np.array([2, 3, 4, 5, 6])
+
+    expected_r2 = 0.5
+    r2 = r2_score(y_true, y_pred)
+
+    assert np.isclose(expected_r2, r2, atol=1e-5, rtol=1e-5)
+
+
+def test_r2_reversed_arrays():
+    y_true = np.array([1, 2, 3, 4, 5])
+    y_pred = np.array([5, 4, 3, 2, 1])
+
+    expected_r2 = -3.0
+    r2 = r2_score(y_true, y_pred)
+
+    assert np.isclose(expected_r2, r2, atol=1e-5, rtol=1e-5)
+
+
+def test_r2_large_numbers():
+    y_true = np.array([10, 20, 30, 40, 50])
+    y_pred = np.array([15, 25, 35, 45, 55])
+
+    expected_r2 = 0.875
+    r2 = r2_score(y_true, y_pred)
+
+    assert np.isclose(expected_r2, r2, atol=1e-5, rtol=1e-5)
+
+
+def test_mse_decimal_numbers():
+    y_true = np.array([0.5, 0.6, 0.7, 0.8, 0.9])
+    y_pred = np.array([0.6, 0.7, 0.8, 0.9, 1.0])
+
+    expected_r2 = 0.5
+    r2 = r2_score(y_true, y_pred)
+
+    assert np.isclose(expected_r2, r2, atol=1e-5, rtol=1e-5)
+
+
+def test_mse_with_outliers():
+    y_true = np.array([10, 20, 30, 40, 135])
+    y_pred = np.array([15, 25, 35, 45, 55])
+
+    expected_r2 = 1 - 6500 / 10180
+    r2 = r2_score(y_true, y_pred)
+
+    assert np.isclose(expected_r2, r2, atol=1e-5, rtol=1e-5)