零售店的需求预测和库存管理 - SARIMA 模型-Python教程-PHP中文网

零售店每天处理大量库存，使得库存监控和管理变得更加繁琐。传统的零售商店库存管理方法繁琐，监控、跟踪和管理效率低下。这就需要一个强大的数字化库存管理系统，该系统可以无缝执行零售商店库存分析，以减少手头库存，并以更少的体力劳动实现更多库存销售。

本文展示了如何使用时间序列机器学习模型 SARIMA 来高效地执行零售商店库存分析，并计算随着时间的推移满足客户需求所需的库存参数，从而使零售商店获得最大利润。

Demand Forecasting and Inventory Management in Retail Store - SARIMA Model

数据集

首先，下载数据集。该数据集包含特定产品的历史记录，包括日期、产品需求和当前库存水平等信息。

代码

执行需求预测和库存管理的Python代码如下。

import pandas as pd
import numpy as np
import plotly.express as px
from statsmodels.graphics.tsaplots import plot_acf, plot_pacf
import matplotlib.pyplot as plt
from statsmodels.tsa.statespace.sarimax import SARIMAX

data = pd.read_csv("demand_inventory.csv")
print(data.head())

data = data.drop(columns=['Unnamed: 0'])

fig_demand = px.line(data, x='Date', y='Demand', title='Demand Over Time')
fig_demand.show()

fig_inventory = px.line(data, x='Date', y='Inventory', title='Inventory Over Time')
fig_inventory.show()

data['Date'] = pd.to_datetime(data['Date'], format='%Y/%m/%d')
time_series = data.set_index('Date')['Demand']

differenced_series = time_series.diff().dropna()

# Plot ACF and PACF of differenced time series
fig, axes = plt.subplots(1, 2, figsize=(12, 4))
plot_acf(differenced_series, ax=axes[0])
plot_pacf(differenced_series, ax=axes[1])
plt.show()

order = (1, 1, 1)
seasonal_order = (1, 1, 1, 2)
model = SARIMAX(time_series, order=order, seasonal_order=seasonal_order)
model_fit = model.fit(disp=False)
future_steps = 10
predictions = model_fit.predict(len(time_series), len(time_series) + future_steps - 1)
predictions = predictions.astype(int)
print(predictions)

# Create date indices for the future predictions
future_dates = pd.date_range(start=time_series.index[-1] + pd.DateOffset(days=1), periods=future_steps, freq='D')

# Create a pandas Series with the predicted values and date indices
forecasted_demand = pd.Series(predictions, index=future_dates)

# Initial inventory level
initial_inventory = 5500

# Lead time (number of days it takes to replenish inventory)
lead_time = 1

# Service level (probability of not stocking out)
service_level = 0.95

# Calculate the optimal order quantity using the Newsvendor formula
z = np.abs(np.percentile(forecasted_demand, 100 * (1 - service_level)))
order_quantity = np.ceil(forecasted_demand.mean() + z).astype(int)

# Calculate the reorder point
reorder_point = forecasted_demand.mean() * lead_time + z

# Calculate the optimal safety stock
safety_stock = reorder_point - forecasted_demand.mean() * lead_time

# Calculate the total cost (holding cost + stockout cost)
holding_cost = 0.1  # it's different for every business, 0.1 is an example
stockout_cost = 10  # # it's different for every business, 10 is an example
total_holding_cost = holding_cost * (initial_inventory + 0.5 * order_quantity)
total_stockout_cost = stockout_cost * np.maximum(0, forecasted_demand.mean() * lead_time - initial_inventory)

# Calculate the total cost
total_cost = total_holding_cost + total_stockout_cost

print("Optimal Order Quantity:", order_quantity)
print("Reorder Point:", reorder_point)
print("Safety Stock:", safety_stock)
print("Total Cost:", total_cost)

登录后复制