Diabetes — логистическая регрессия и coef_

ДЛЯ НОВИЧКОВ

Бинарная классификация: есть диабет (Outcome = 1) или нет (0). Задача из учебных курсов ML — после Insurance и перед Titanic.

Зависимости: pip install pandas scikit-learn tensorflow

1. Загрузка

import pandas as pd

cols = [
    "Pregnancies", "Glucose", "BloodPressure", "SkinThickness",
    "Insulin", "BMI", "DiabetesPedigreeFunction", "Age", "Outcome",
]
url = "https://raw.githubusercontent.com/jbrownlee/Datasets/master/pima-indians-diabetes.data.csv"
df = pd.read_csv(url, names=cols)
print(df["Outcome"].value_counts())

2. Логистическая регрессия

from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from sklearn.pipeline import Pipeline
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import accuracy_score, confusion_matrix, classification_report

X = df.drop(columns="Outcome")
y = df["Outcome"]

X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.2, random_state=42, stratify=y
)

pipe = Pipeline([
    ("scaler", StandardScaler()),
    ("clf", LogisticRegression(max_iter=1000)),
])
pipe.fit(X_train, y_train)
y_pred = pipe.predict(X_test)

print(classification_report(y_test, y_pred))
print("Confusion matrix:\n", confusion_matrix(y_test, y_pred))

clf = pipe.named_steps["clf"]
for name, coef in zip(X.columns, clf.coef_[0]):
    print(f"{name:28s} {coef:+.4f}")

Самопроверка: какой признак по модулю coef_ сильнее всего связан с диабетом? Часто это Glucose и BMI.

3. Нейросеть (Keras) — сравнение

import os
os.environ.setdefault("TF_CPP_MIN_LOG_LEVEL", "2")
import tensorflow as tf
import numpy as np

scaler = StandardScaler()
X_tr = scaler.fit_transform(X_train).astype("float32")
X_te = scaler.transform(X_test).astype("float32")

nn = tf.keras.Sequential([
    tf.keras.layers.Input(shape=(X_tr.shape[1],)),
    tf.keras.layers.Dense(16, activation="relu"),
    tf.keras.layers.Dense(8, activation="relu"),
    tf.keras.layers.Dense(1, activation="sigmoid"),
])
nn.compile(optimizer="adam", loss="binary_crossentropy", metrics=["accuracy"])
nn.fit(X_tr, y_train, epochs=50, batch_size=32, verbose=0)

y_nn = (nn.predict(X_te, verbose=0).ravel() >= 0.5).astype(int)
print("Logistic accuracy:", accuracy_score(y_test, y_pred))
print("Keras NN accuracy:", accuracy_score(y_test, y_nn))

На малых табличных данных NN не всегда лучше логистической регрессии — это нормальный результат.

4. Дерево решений (опционально)

from sklearn.tree import DecisionTreeClassifier

tree = DecisionTreeClassifier(max_depth=4, random_state=42)
tree.fit(X_train, y_train)
print("Tree accuracy:", accuracy_score(y_test, tree.predict(X_test)))

Визуализация — plot_tree.

Дальше