Ravindra BagaleCourses & study guides मराठी Track your progress

65. How to Forecast Sales with Time Series in Python (pandas and sklearn)

65.4 Moving Average Forecast

By , trainer LinkedIn Last updated:

Why. Last month can be a spike. A short average smooths noise. Banks and clinics use this as a calm baseline.

What. A moving average (MA) of the last k points. For a flat forecast, use the MA of the last k train points for every future month (simple teaching version).

How.

MA(3) flat forecast

  1. Take the last 3 train values.
  2. Compute their mean.
  3. Use that mean as the prediction for each test month.
  4. Later you can also roll the window forward one step at a time.

Worked example A — MA(3) on Mauli train tail.

import pandas as pd
import numpy as np

df = pd.read_csv("mauli_monthly_sales.csv", parse_dates=["month"]).sort_values("month")
train = df.iloc[:30]
test = df.iloc[30:]
ma3 = float(train["sales_inr"].iloc[-3:].mean())
pred = np.full(len(test), int(round(ma3)))
print("Last 3 train values:", train["sales_inr"].iloc[-3:].tolist())
print("MA3:", round(ma3, 2))
print(list(zip(test["month"].dt.strftime("%Y-%m"), pred)))

Output you should see.

Last 3 train values: [539600, 594100, 631700]
MA3: 588466.67
[('2026-05', 588467), ('2026-06', 588467), ('2026-07', 588467), ('2026-08', 588467), ('2026-09', 588467), ('2026-10', 588467)]

What it means. Smoother than last-value (631700), but still flat — it cannot rise into festival months by itself.

Worked example B — US bank deposits ($).

Dataset note: fictional Hilltop Bank Austin branch new deposits. 8 rows. Columns: month, new_deposits_usd.

import pandas as pd

bank = pd.DataFrame({
    "month": pd.date_range("2025-01-01", periods=8, freq="MS"),
    "new_deposits_usd": [1200000, 1150000, 1300000, 1280000, 1400000, 1350000, 1500000, 1480000],
})
bank["ma3"] = bank["new_deposits_usd"].rolling(3).mean().round(0)
print(bank.to_string(index=False))

Output you should see.

     month  new_deposits_usd       ma3
2025-01-01           1200000       NaN
2025-02-01           1150000       NaN
2025-03-01           1300000 1216667.0
...
2025-08-01           1480000 1443333.0

What it means. First two MA3 cells are NaN (need 3 points). August MA3 ≈ $1.44M — a calm view of recent deposits.

Worked example C — When MA loses.

If sales jump every October, a flat MA under-shoots. Seasonal naive or lag-12 features handle that better.

Ravindra Bagale's Tip

MA is a calm baseline. If festivals jump, do not trust MA alone — keep a seasonal method.