mex-p3.py

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns

from sklearn.cluster import KMeans
from sklearn.preprocessing import StandardScaler

from google.colab import drive
drive.mount('/content/drive')


# Carregar os dados
file_path = "/content/drive/MyDrive/mex-data/Online Retail.xlsx"
df = pd.read_excel(file_path)

# Remover valores nulos
df = df.dropna()

# Remover valores negativos
df = df[(df["Quantity"] > 0) & (df["UnitPrice"] > 0)]

# Criar a feature de valor total gasto
df["TotalPrice"] = df["Quantity"] * df["UnitPrice"]

# Agrupar por Cliente e calcular métricas relevantes
customer_data = df.groupby("CustomerID").agg({
    "InvoiceNo": "count",  # Número de compras
    "Quantity": "sum",      # Quantidade total comprada
    "TotalPrice": "sum"     # Total gasto
}).rename(columns={"InvoiceNo": "NumCompras"})

# Normalizar os dados
scaler = StandardScaler()
customer_data_scaled = scaler.fit_transform(customer_data)

# Definir o número de clusters
k = 3  # Ajuste conforme necessário

# Criar e treinar o modelo K-Means
kmeans = KMeans(n_clusters=k, random_state=42, n_init=10)
customer_data["Cluster"] = kmeans.fit_predict(customer_data_scaled)

# Visualizar os clusters
print(customer_data.head())

# Gráfico de dispersão dos clusters
plt.figure(figsize=(10, 6))
sns.scatterplot(
    x=customer_data["TotalPrice"],
    y=customer_data["NumCompras"],
    hue=customer_data["Cluster"],
    palette="viridis"
)

plt.xlabel("Total Gasto")
plt.ylabel("Número de Compras")
plt.title("Segmentação de Clientes")
plt.show()