Time Series Forecast with Uncertainty Band in Pygal (Python)

The same plot in 14 other libraries — Python: Altair, Bokeh, lets-plot, Matplotlib, Plotly, plotnine, Seaborn; R: ggplot2; Julia: Makie.jl; JavaScript: Chart.js, D3.js, Apache ECharts, Highcharts, MUI X Charts. Compare all 15 side by side: Time Series Forecast with Uncertainty Band in Python, R, Julia and JavaScript.

A time series plot that displays historical observed data followed by a forecast projection with confidence intervals or uncertainty bands. The plot clearly distinguishes between the historical period and the forecast period using a vertical line marker, with shaded bands representing different confidence levels (typically 80% and 95%). This visualization is essential for communicating prediction uncertainty in forecasting applications, helping stakeholders understand both the expected values and the range of possible outcomes.

Time Series Forecast with Uncertainty Band rendered with Pygal

Renders

Python source (Pygal)

""" anyplot.ai
timeseries-forecast-uncertainty: Time Series Forecast with Uncertainty Band
Library: pygal 3.1.3 | Python 3.14.10
Quality: 88/100 | Updated: 2026-11-08
"""

import os
import sys
sys.path[:] = [p for p in sys.path if p not in ("", ".", os.path.dirname(__file__))]

import numpy as np
import pandas as pd
import pygal
from pygal.style import Style


THEME = os.getenv("ANYPLOT_THEME", "light")
PAGE_BG = "#FAF8F1" if THEME == "light" else "#1A1A17"
INK = "#1A1A17" if THEME == "light" else "#F0EFE8"
INK_SOFT = "#4A4A44" if THEME == "light" else "#B8B7B0"
RULE = "rgba(26,26,23,0.15)" if THEME == "light" else "rgba(240,239,232,0.15)"

BRAND = "#009E73"  # historical observations
FORECAST_COLOR = "#C475FD"  # forecast line; its bands are translucent tints of the same lavender
BAND_95 = "rgba(196, 117, 253, 0.20)" if THEME == "light" else "rgba(196, 117, 253, 0.22)"
BAND_80 = "rgba(196, 117, 253, 0.45)" if THEME == "light" else "rgba(196, 117, 253, 0.42)"

# Constants
TICK_COUNT = 7  # about this many date ticks along the x axis

# Columns: set these to your table's column names
DATE = "Month"  # role: date
ACTUAL = "Monthly sales ($)"  # role: actual
FORECAST = "Forecast"  # role: forecast
LOWER_80 = "Lower 80%"  # role: lower_80
UPPER_80 = "Upper 80%"  # role: upper_80
LOWER_95 = "Lower 95%"  # role: lower_95
UPPER_95 = "Upper 95%"  # role: upper_95

# Data: replace this block with your own table, for example df = pd.read_csv("forecast.csv")
np.random.seed(42)
months = 42
split_idx = 36

t = np.arange(split_idx)
actual = 1000 + 50 * t + 100 * np.sin(2 * np.pi * t / 12) + np.random.normal(0, 30, split_idx)

t_forecast = np.arange(split_idx, months)
forecast_values = actual[-1] + 30 * (t_forecast - split_idx) + 50 * np.sin(2 * np.pi * t_forecast / 12)

history_gap = np.full(split_idx, np.nan)
forecast_gap = np.full(months - split_idx, np.nan)
df = pd.DataFrame(
    {
        DATE: pd.date_range("2022-01-01", periods=months, freq="MS"),
        ACTUAL: np.concatenate([actual, forecast_gap]),
        FORECAST: np.concatenate([history_gap, forecast_values]),
        LOWER_80: np.concatenate([history_gap, forecast_values - 100]),
        UPPER_80: np.concatenate([history_gap, forecast_values + 100]),
        LOWER_95: np.concatenate([history_gap, forecast_values - 150]),
        UPPER_95: np.concatenate([history_gap, forecast_values + 150]),
    }
)

# Plot: rows numbered from 1 (no tick at 0, so pygal draws its left spine at the edge)
df = df.sort_values(DATE).reset_index(drop=True)
df["x"] = np.arange(1, len(df) + 1, dtype=float)
hist = df[df[ACTUAL].notna()]
fc = df[df[FORECAST].notna() & df[ACTUAL].isna()]
bands = fc.dropna(subset=[LOWER_80, UPPER_80, LOWER_95, UPPER_95])
origin = (float(hist["x"].iloc[-1]), float(hist[ACTUAL].iloc[-1]))  # forecast and bands start at the last observation

# Bands as polygons between their bounds (upper bound forward, lower bound back), starting at the origin:
# the 80% band, and the two strips of the 95% band outside it, so each colour has exactly one region
fc_x = [origin[0], *bands["x"]]
band_80 = list(zip(fc_x, [origin[1], *bands[UPPER_80]], strict=True)) + list(
    zip(fc_x[::-1], [*bands[LOWER_80][::-1], origin[1]], strict=True)
)
band_95_upper = list(zip(fc_x, [origin[1], *bands[UPPER_95]], strict=True)) + list(
    zip(fc_x[::-1], [*bands[UPPER_80][::-1], origin[1]], strict=True)
)
band_95_lower = list(zip(fc_x, [origin[1], *bands[LOWER_80]], strict=True)) + list(
    zip(fc_x[::-1], [*bands[LOWER_95][::-1], origin[1]], strict=True)
)

# A 1-2-5 tick step giving about ten y ticks across the data and both 95% bounds
data_lo = np.nanmin([df[ACTUAL].min(), df[LOWER_95].min()])
data_hi = np.nanmax([df[ACTUAL].max(), df[UPPER_95].max()])
raw_step = (data_hi - data_lo) / 10
magnitude = 10 ** np.floor(np.log10(raw_step))
STEP = magnitude * next(m for m in (1, 2, 5, 10) if m * magnitude >= raw_step)
y_lo = np.floor(data_lo / STEP) * STEP
y_hi = np.ceil(data_hi / STEP) * STEP + STEP / 2
# The forecast-start line and its caption are drawn only when the table holds a forecast period
forecast_start = float(fc["x"].iloc[0]) if len(fc) else None
marker = [(forecast_start, y_lo), (forecast_start, y_hi)] if len(fc) else []
caption = [{"value": (forecast_start + 0.15, y_hi - STEP / 4), "label": "Forecast →"}] if len(fc) else []
tick_every = max(1, round(len(df) / TICK_COUNT))

LABEL_FS = 40
custom_style = Style(
    background=PAGE_BG,
    plot_background=PAGE_BG,
    foreground=INK,
    foreground_strong=INK,
    foreground_subtle=INK_SOFT,
    colors=(INK, BAND_95, BAND_95, BAND_80, INK_SOFT, BRAND, FORECAST_COLOR),
    value_colors=(INK,) * 7,
    opacity=1,
    opacity_hover=1,
    stroke_opacity=1,
    stroke_opacity_hover=1,
    stroke_width=5,
    guide_stroke_color=RULE,
    guide_stroke_dasharray="none",
    major_guide_stroke_color=RULE,
    major_guide_stroke_dasharray="none",
    title_font_size=60,
    label_font_size=44,
    major_label_font_size=44,
    legend_font_size=42,
    # pygal prints a point's label value_font_size px right of it and 4/3 of that below it;
    # at 0.26 x the label size the text sits vertically centred on its point
    value_font_size=round(0.26 * LABEL_FS),
    value_label_font_size=LABEL_FS,
    tooltip_font_size=34,
)

chart = pygal.XY(
    width=3200,
    height=1800,
    style=custom_style,
    title="timeseries-forecast-uncertainty · python · pygal · anyplot.ai",
    x_title=DATE,
    y_title=ACTUAL,
    show_x_guides=False,
    show_y_guides=True,
    show_legend=True,
    legend_at_bottom=True,
    legend_at_bottom_columns=5,  # one row; pygal counts the untitled series too when it reserves legend rows
    legend_box_size=36,
    truncate_legend=-1,
    xrange=(float(df["x"].iloc[0]), float(df["x"].iloc[-1]) + 0.5),
    range=(float(y_lo), float(y_hi)),
    print_labels=True,
    print_values=False,
    margin=40,
    spacing=24,
    css=(
        "file://style.css",
        "file://graph.css",
        "inline:.serie-1 .line, .serie-2 .line, .serie-3 .line { stroke-width: 0 !important; }",
        "inline:.serie-4 .line { stroke-width: 3px !important; }",
        "inline:.text-overlay text.label { font-weight: bold; }",
        "inline:.plot_title { font-weight: bold; }",
        # pygal sets legend text only 5 px right of its swatch; push it clear of the box
        "inline:.legends .legend text { transform: translate(14px, 0px); }",
        # The dashed series get dashed outline swatches instead of solid squares
        "inline:#activate-serie-4 rect, #activate-serie-6 rect "
        "{ fill: none !important; stroke-width: 5px; stroke-dasharray: 9, 5; }",
        f"inline:#activate-serie-4 rect {{ stroke: {INK_SOFT} !important; }}",
        f"inline:#activate-serie-6 rect {{ stroke: {FORECAST_COLOR} !important; }}",
    ),
)
chart.x_labels = [
    {"value": int(row["x"]), "label": row[DATE].strftime("%b %y")} for _, row in df.iloc[::tick_every].iterrows()
]

# Label first and bands next, so the lines are drawn on top of the fills
chart.add(None, caption, stroke=False, dots_size=0)
chart.add("95% interval", band_95_upper, fill=True, show_dots=False)
chart.add(None, band_95_lower, fill=True, show_dots=False)
chart.add("80% interval", band_80, fill=True, show_dots=False)
chart.add("Forecast start", marker, show_dots=False, stroke_style={"dasharray": "14, 10"})
chart.add(
    "Observed history",
    list(zip(hist["x"], hist[ACTUAL], strict=True)),
    dots_size=5,
    formatter=lambda p: f"${p[1]:,.0f}",
)
chart.add(
    FORECAST,
    [origin, *zip(fc["x"], fc[FORECAST], strict=True)],
    dots_size=5,
    stroke_style={"dasharray": "22, 12"},
    formatter=lambda p: f"${p[1]:,.0f}",
)

# Save
chart.render_to_png(f"plot-{THEME}.png")
with open(f"plot-{THEME}.html", "wb") as f:
    f.write(chart.render())

Retrieve this implementation

Runnable source as JSON, for any HTTP client: https://api.anyplot.ai/specs/timeseries-forecast-uncertainty/pygal/code. Any spec id and library id listed in llms-full.txt fit the same URL shape; every URL below is complete and callable.

{
  "spec_id": "timeseries-forecast-uncertainty",
  "language": "python",
  "library": "pygal",
  "page": "https://anyplot.ai/timeseries-forecast-uncertainty/python/pygal",
  "hub": "https://anyplot.ai/timeseries-forecast-uncertainty",
  "code_json": "https://api.anyplot.ai/specs/timeseries-forecast-uncertainty/pygal/code",
  "spec_json": "https://api.anyplot.ai/specs/timeseries-forecast-uncertainty",
  "render_light_png": "https://storage.googleapis.com/anyplot-images/plots/timeseries-forecast-uncertainty/python/pygal/plot-light.png",
  "render_dark_png": "https://storage.googleapis.com/anyplot-images/plots/timeseries-forecast-uncertainty/python/pygal/plot-dark.png",
  "interactive_light_html": "https://storage.googleapis.com/anyplot-images/plots/timeseries-forecast-uncertainty/python/pygal/plot-light.html",
  "interactive_dark_html": "https://storage.googleapis.com/anyplot-images/plots/timeseries-forecast-uncertainty/python/pygal/plot-dark.html",
  "quality_score": 88.0,
  "license": "MIT",
  "guide": "https://anyplot.ai/llms.txt"
}

Part of Time Series Forecast with Uncertainty Band on anyplot.ai.

Other implementations