如何用Matplotlib平滑趋势线？附现有实现代码

Hey there! Let's tackle that jagged line problem you're facing with your plot. First, let's fix a small but important oversight in your current code: when you convert the W column to datetime, you're not assigning the result back to df['W']. That means your x-axis is still being treated as string values instead of proper datetime, which can mess up plotting logic. Let's correct that first, then dive into practical smoothing techniques.

Step 1: Fix the Datetime Conversion #

First, update this line to ensure your datetime column is properly set:

df['W'] = pd.to_datetime(df['W'])  # Assign the converted datetime back to df['W']

Now, let's look at three effective ways to smooth your trend lines, each with its own use case:

1. Rolling Average (Simple Moving Average) #

This is a straightforward method for time-series data—it averages values over a sliding window to reduce noise. Perfect if you want a simple, interpretable smooth line that retains the general trend.

Here's how to implement it:

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt

# Generate and prep data
df = pd.DataFrame(data=np.random.randn(5,4), index=['A','B','C','D','E'], columns=['W','X','Y','Z'])
df['W'] = pd.to_datetime(['10/01/2018 12:00:00','10/03/2018 13:00:00', '10/03/2018 12:30:00','10/04/2018 12:05:00', '10/08/2018 12:00:15'])
# Sort by datetime first (critical for rolling window logic!)
df = df.sort_values('W')

# Calculate rolling averages (window=2 means average 2 consecutive points)
df['X_smoothed'] = df['X'].rolling(window=2, min_periods=1).mean()
df['Y_smoothed'] = df['Y'].rolling(window=2, min_periods=1).mean()

# Plot original and smoothed lines
fig, ax = plt.subplots()
# Original lines (lighter, transparent to highlight smoothed versions)
df.plot(x="W", y="X", ax=ax, color='salmon', alpha=0.5, marker='o', label='Original X')
df.plot(x="W", y="Y", ax=ax, color='royalblue', alpha=0.4, marker='o', label='Original Y')
# Smoothed lines (bold, solid)
df.plot(x="W", y="X_smoothed", ax=ax, color='salmon', linewidth=2, label='Smoothed X')
df.plot(x="W", y="Y_smoothed", ax=ax, color='royalblue', linewidth=2, label='Smoothed Y')

ax.legend()
plt.xticks(rotation=45)
plt.show()

Pro tip: Adjust the window parameter based on your data—larger windows mean more smoothing but less responsiveness to sudden changes.

2. LOESS/LOWESS Smoothing #

LOESS (Locally Estimated Scatterplot Smoothing) fits a local regression to small subsets of your data, creating a flexible, non-linear smooth line. It's ideal for capturing complex trends without overfitting to noise.

You'll need the statsmodels library first—install it with pip install statsmodels:

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from statsmodels.nonparametric.smoothers_lowess import lowess

# Generate and prep data
df = pd.DataFrame(data=np.random.randn(5,4), index=['A','B','C','D','E'], columns=['W','X','Y','Z'])
df['W'] = pd.to_datetime(['10/01/2018 12:00:00','10/03/2018 13:00:00', '10/03/2018 12:30:00','10/04/2018 12:05:00', '10/08/2018 12:00:15'])
df = df.sort_values('W')

# Convert datetime to numeric (LOESS works with numerical inputs)
df['W_numeric'] = df['W'].astype('int64') / 10**9  # Convert to seconds since epoch

# Apply LOESS smoothing (frac controls smoothing strength)
x_smoothed = lowess(df['X'], df['W_numeric'], frac=0.5)[:, 1]
y_smoothed = lowess(df['Y'], df['W_numeric'], frac=0.5)[:, 1]

# Add smoothed values back to dataframe
df['X_smoothed'] = x_smoothed
df['Y_smoothed'] = y_smoothed

# Plot
fig, ax = plt.subplots()
df.plot(x="W", y="X", ax=ax, color='salmon', alpha=0.5, marker='o', label='Original X')
df.plot(x="W", y="Y", ax=ax, color='royalblue', alpha=0.4, marker='o', label='Original Y')
df.plot(x="W", y="X_smoothed", ax=ax, color='salmon', linewidth=2, label='Smoothed X')
df.plot(x="W", y="Y_smoothed", ax=ax, color='royalblue', linewidth=2, label='Smoothed Y')

ax.legend()
plt.xticks(rotation=45)
plt.show()

Note: The frac parameter sets the smoothing strength—higher values (e.g., 0.7) mean more smoothing, lower values (e.g., 0.3) mean the line stays closer to your original data.

3. Spline Interpolation #

Spline interpolation creates a smooth, continuous curve that passes through every one of your data points. It's great if you want a polished line that preserves the exact shape of your data while eliminating jagged edges.

Use scipy.interpolate for this—install scipy with pip install scipy:

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from scipy.interpolate import make_interp_spline

# Generate and prep data
df = pd.DataFrame(data=np.random.randn(5,4), index=['A','B','C','D','E'], columns=['W','X','Y','Z'])
df['W'] = pd.to_datetime(['10/01/2018 12:00:00','10/03/2018 13:00:00', '10/03/2018 12:30:00','10/04/2018 12:05:00', '10/08/2018 12:00:15'])
df = df.sort_values('W')

# Convert datetime to numeric
df['W_numeric'] = df['W'].astype('int64') / 10**9

# Create a denser x-axis for a smoother curve
w_dense = np.linspace(df['W_numeric'].min(), df['W_numeric'].max(), 100)

# Generate spline curves
spline_x = make_interp_spline(df['W_numeric'], df['X'])
x_smoothed = spline_x(w_dense)

spline_y = make_interp_spline(df['W_numeric'], df['Y'])
y_smoothed = spline_y(w_dense)

# Convert back to datetime for plotting
w_dense_datetime = pd.to_datetime(w_dense, unit='s')

# Plot
fig, ax = plt.subplots()
# Original data points
ax.plot(df['W'], df['X'], color='salmon', alpha=0.5, marker='o', label='Original X')
ax.plot(df['W'], df['Y'], color='royalblue', alpha=0.4, marker='o', label='Original Y')
# Smoothed spline curves
ax.plot(w_dense_datetime, x_smoothed, color='salmon', linewidth=2, label='Smoothed X')
ax.plot(w_dense_datetime, y_smoothed, color='royalblue', linewidth=2, label='Smoothed Y')

ax.legend()
plt.xticks(rotation=45)
plt.show()

Bonus: The 100 in np.linspace controls how dense the smooth curve is—higher numbers mean even smoother lines.

内容的提问来源于stack exchange，提问作者user9185511