diff --git a/TODO.md b/TODO.md
index b710c5e..65a481b 100644
--- a/TODO.md
+++ b/TODO.md
@@ -207,6 +207,12 @@ After completing a milestone, create a pull request with your changes for review
 - [x] Add global theme toggle stored in session state
 - [x] Update tests to verify theme CSS output
 
+## PR21: Forecasting Models
+
+- [x] Create utils/time_series.py with ARIMA and naive forecast functions
+- [x] Add UI controls on Time Series page for model selection and forecast horizon
+- [x] Write tests for forecasting utilities
+
 ## Notes for Development
 
 - Create comprehensive commit messages that clearly describe changes
diff --git a/pages/time_series.py b/pages/time_series.py
index 30838ab..66ecbec 100644
--- a/pages/time_series.py
+++ b/pages/time_series.py
@@ -7,6 +7,7 @@
 import streamlit as st
 
 from utils import ui, viz
+from utils import time_series as ts
 from utils.logging import configure_logging
 
 configure_logging()
@@ -38,6 +39,12 @@ def main() -> None:
         period = st.number_input(
             "Seasonal Period", min_value=2, value=2, step=1, key="ts_period"
         )
+        model_choice = st.selectbox(
+            "Forecast Model", ["Naive", "ARIMA"], key="ts_model"
+        )
+        horizon = st.number_input(
+            "Forecast Horizon", min_value=1, value=5, step=1, key="ts_horizon"
+        )
 
     if st.button("Generate Plots"):
         ts_fig = viz.time_series_plot(df, time_col, value_col, title="Time Series")
@@ -52,6 +59,20 @@ def main() -> None:
                 mime=f"image/{export_fmt}",
             )
 
+        series = df.set_index(time_col)[value_col]
+        if model_choice == "ARIMA":
+            try:
+                forecast = ts.arima_forecast(series, steps=horizon)
+            except ImportError:
+                st.error("statsmodels is required for ARIMA forecasting.")
+                forecast = None
+        else:
+            forecast = ts.naive_forecast(series, steps=horizon)
+
+        if forecast is not None:
+            st.subheader("Forecast")
+            st.write(forecast.to_frame(name="forecast"))
+
         dec_fig = viz.decomposition_plot(
             df.set_index(time_col)[value_col], period=period, title="Decomposition"
         )
diff --git a/tests/test_pages.py b/tests/test_pages.py
index b1235a8..9054851 100644
--- a/tests/test_pages.py
+++ b/tests/test_pages.py
@@ -112,6 +112,13 @@ def test_time_series_page_contents():
     assert "decomposition_plot" in content
 
 
+def test_time_series_page_forecast_widgets():
+    with open("pages/time_series.py", "r", encoding="utf-8") as f:
+        content = f.read()
+    assert "Forecast Model" in content
+    assert "Forecast Horizon" in content
+
+
 def test_datetime_cols_persist_after_transforms():
     import streamlit as st
     from utils import transform, eda
diff --git a/tests/test_time_series_utils.py b/tests/test_time_series_utils.py
new file mode 100644
index 0000000..0bbb78e
--- /dev/null
+++ b/tests/test_time_series_utils.py
@@ -0,0 +1,29 @@
+import pandas as pd
+import pytest
+
+from utils import time_series
+
+
+def sample_series():
+    return pd.Series(
+        range(5), index=pd.date_range("2021-01-01", periods=5, freq="D")
+    )
+
+
+def test_naive_forecast_extends_index():
+    series = sample_series()
+    forecast = time_series.naive_forecast(series, steps=3)
+    assert len(forecast) == 3
+    assert forecast.iloc[0] == series.iloc[-1]
+    assert forecast.index[0] == series.index[-1] + series.index.freq
+
+
+def test_arima_forecast_runs_or_errors():
+    series = sample_series()
+    try:
+        fc = time_series.arima_forecast(series, steps=2)
+        assert len(fc) == 2
+    except ImportError:
+        pytest.skip("statsmodels not available")
+
+
diff --git a/utils/time_series.py b/utils/time_series.py
new file mode 100644
index 0000000..4703768
--- /dev/null
+++ b/utils/time_series.py
@@ -0,0 +1,46 @@
+"""Time series forecasting utilities."""
+
+from __future__ import annotations
+
+from typing import Tuple
+
+import pandas as pd
+
+
+def _extend_index(index: pd.Index, steps: int) -> pd.Index:
+    """Return an extended index for forecast values."""
+    if isinstance(index, pd.DatetimeIndex) and index.freq is not None:
+        start = index[-1] + index.freq
+        return pd.date_range(start, periods=steps, freq=index.freq)
+    return pd.RangeIndex(index[-1] + 1, index[-1] + 1 + steps)
+
+
+def naive_forecast(series: pd.Series, steps: int = 1) -> pd.Series:
+    """Forecast future values using the last observed value."""
+    last = series.iloc[-1]
+    index = _extend_index(series.index, steps)
+    return pd.Series([last] * steps, index=index, name="naive_forecast")
+
+
+def arima_forecast(
+    series: pd.Series,
+    *,
+    order: Tuple[int, int, int] = (1, 1, 0),
+    steps: int = 1,
+) -> pd.Series:
+    """Forecast future values using an ARIMA model.
+
+    Requires the ``statsmodels`` package. If it is not installed an
+    ``ImportError`` is raised.
+    """
+    try:
+        from statsmodels.tsa.arima.model import ARIMA
+    except Exception as exc:  # pragma: no cover - optional dependency
+        raise ImportError("statsmodels is required for ARIMA forecasting") from exc
+
+    model = ARIMA(series, order=order)
+    fitted = model.fit()
+    forecast = fitted.forecast(steps=steps)
+    index = _extend_index(series.index, steps)
+    forecast.index = index
+    return forecast