Spaces:
Sleeping
Sleeping
File size: 5,315 Bytes
f73646a | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 | import pandas as pd
import numpy as np
from sklearn.feature_selection import chi2, f_classif, mutual_info_classif, mutual_info_regression
#from utils.decision import decide_pipeline
def detect_target_type(y):
# target detection ... categorical (boolean , object , categorical numbers )
if y.dtype == "object":
return "categorical"
if str(y.dtype) == "category":
return "categorical"
if str(y.dtype) == "bool":
return "categorical"
if pd.api.types.is_numeric_dtype(y):
unique_vals = y.nunique()
# binary or multiclass target
if unique_vals <= 10:
return "categorical"
return "numerical"
return "categorical"
def clean_numeric_scores(series):
# remove infinity , nan ,
# sort descending
series = series.replace([np.inf, -np.inf], np.nan).dropna()
return series.sort_values(ascending=False)
def feature_selection(df, target ,decisions ):
results = {}
if target not in df.columns:
return results
# separate columns
numeric_cols = df.select_dtypes(include=["int64", "float64"]).columns.tolist()
categorical_cols = df.select_dtypes(include=["object", "category", "bool"]).columns.tolist()
# remove target
if target in numeric_cols:
numeric_cols.remove(target)
if target in categorical_cols:
categorical_cols.remove(target)
# remove high cardinality categorical columns
categorical_cols = [
col for col in categorical_cols
if df[col].nunique(dropna=True) <= 50
# cardh
]
y = df[target].copy()
target_type = detect_target_type(y)
# case 1 -> numerical target (regression)
if target_type == "numerical":
# numerical Features vs target
if len(numeric_cols) > 0:
X_num = df[numeric_cols]
corr_scores = X_num.corrwith(y).abs()
corr_scores = clean_numeric_scores(corr_scores)
if len(corr_scores) > 0:
results["numerical_correlation"] = corr_scores
# mutual information
try:
mi = mutual_info_regression(
X_num.fillna(X_num.median()),
y
)
mi_scores = pd.Series(
mi,
index=numeric_cols
)
mi_scores = clean_numeric_scores(mi_scores)
if len(mi_scores) > 0:
results["numerical_mutual_info"] = mi_scores
except:
pass
# categorical features vs target
if len(categorical_cols) > 0:
X_cat = pd.get_dummies(
df[categorical_cols],
drop_first=True
)
if X_cat.shape[1] > 0:
try:
f_scores, _ = f_classif(X_cat, y)
anova_scores = pd.Series(
f_scores,
index=X_cat.columns
)
anova_scores = clean_numeric_scores(anova_scores)
if len(anova_scores) > 0:
results["categorical_anova"] = anova_scores
except:
pass
#case 2 -? categorical target (classification)
else:
y_encoded = pd.factorize(y)[0]
# numeric features vs target
if len(numeric_cols) > 0:
X_num = df[numeric_cols].copy()
# fill nulls
for col in X_num.columns:
X_num[col] = X_num[col].fillna(X_num[col].median())
try:
f_scores, _ = f_classif(X_num, y_encoded)
anova_scores = pd.Series(
f_scores,
index=numeric_cols
)
anova_scores = clean_numeric_scores(anova_scores)
if len(anova_scores) > 0:
results["numerical_anova"] = anova_scores
except:
pass
# mutual information ?
try:
mi = mutual_info_classif(X_num, y_encoded)
mi_scores = pd.Series(
mi,
index=numeric_cols
)
mi_scores = clean_numeric_scores(mi_scores)
if len(mi_scores) > 0:
results["numerical_mutual_info"] = mi_scores
except:
pass
# categorical features vs target
if len(categorical_cols) > 0:
X_cat = pd.get_dummies(
df[categorical_cols],
drop_first=True
)
if X_cat.shape[1] > 0:
try:
chi_scores, _ = chi2(X_cat, y_encoded)
chi_scores = pd.Series(
chi_scores,
index=X_cat.columns
)
chi_scores = clean_numeric_scores(chi_scores)
if len(chi_scores) > 0:
results["categorical_chi2"] = chi_scores
except:
pass
return results |