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

65.3 Naive Last-Value and Seasonal Naive Forecasts

65.3 Naive last-value आणि seasonal naive

By Ravindra Bagale, trainer · LinkedIn · Last updated: 31 October 2026

sklearn आधी दोन प्रामाणिक baselines. Fancy model “last month” किंवा “last year same month” पेक्षा वाईट असेल तर काम बाकी आहे.

हे का शिकायचं?

Why. Before sklearn, show two honest baselines. If your fancy model loses to “last month” or “same month last year”, you are not done.

हे काय आहे?

What.

• Naive last-value — next value = last observed value.
• Seasonal naive — next value = value from the same season last cycle (month − 12 for monthly data).

कसं करायचं?

How.

Worked example A — Tiny naive (easiest).

import pandas as pd

tiny = pd.DataFrame({
    "month": ["2026-05", "2026-06", "2026-07", "2026-08"],
    "sales_inr": [510000, 495000, 502000, 540000],
})
print(tiny)
print("Naive forecast for 2026-09:", int(tiny["sales_inr"].iloc[-1]))

Output you should see.

    month  sales_inr
0 2026-05     510000
1 2026-06     495000
2 2026-07     502000
3 2026-08     540000
Naive forecast for 2026-09: 540000

याचा अर्थ. September guess = August actual. Simple. Often surprisingly hard to beat on short horizons.

Worked example B — Mauli train/test naive + seasonal naive.

We hold out May–Oct 2026 (6 months). Train ends Apr 2026. Last train sales = ₹631,700.

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:]
naive = np.full(len(test), int(train["sales_inr"].iloc[-1]))
seasonal = [int(df["sales_inr"].iloc[30 + i - 12]) for i in range(len(test))]
out = pd.DataFrame({
    "month": test["month"].dt.strftime("%Y-%m"),
    "actual": test["sales_inr"].values,
    "naive_pred": naive,
    "seasonal_naive": seasonal,
})
print(out.to_string(index=False))

Output you should see.

  month  actual  naive_pred  seasonal_naive
2026-05  674800      631700          565000
2026-06  620700      631700          532500
2026-07  602500      631700          517600
2026-08  694700      631700          595200
2026-09  799900      631700          693500
2026-10  878700      631700          744300

याचा अर्थ. Naive is a flat line. Seasonal naive follows last year’s shape but sits low when the trend has grown.

Worked example C — Logistics weekly boxes (Bhiwandi).

Dataset note: fictional RapidBox hub weekly shipments. 8 rows. Columns: week_start, boxes_shipped. Currency context: ops KPI (count), not money.

import pandas as pd

logi = pd.DataFrame({
    "week_start": pd.date_range("2026-08-03", periods=8, freq="W-MON"),
    "boxes_shipped": [5200, 5400, 5100, 5800, 6000, 5900, 6100, 6300],
})
print(logi)
print("Naive next week:", int(logi["boxes_shipped"].iloc[-1]))

Output you should see.

  week_start  boxes_shipped
0 2026-08-03           5200
...
7 2026-09-21           6300
Naive next week: 6300